Article Stroke genetics informs drug discovery and risk prediction across ancestries ­­­ https://doi.org/10.1038/s41586-022-05165-3 Previous genome-wide association studies (GWASs) of stroke — the second leading Received: 15 December 2021 cause of death worldwide — were conducted predominantly in populations of European ancestry1,2. Here, in cross-ancestry GWAS meta-analyses of 110,182 patients Accepted: 29 July 2022 who have had a stroke (five ancestries, 33% non-European) and 1,503,898 control Published online: 30 September 2022 individuals, we identify association signals for stroke and its subtypes at 89 (61 new) Open access independent loci: 60 in primary inverse-variance-weighted analyses and 29 in Check for updates secondary meta-regression and multitrait analyses. On the basis of internal cross-ancestry validation and an independent follow-up in 89,084 additional cases of stroke (30% non-European) and 1,013,843 control individuals, 87% of the primary stroke risk loci and 60% of the secondary stroke risk loci were replicated (P < 0.05). Effect sizes were highly correlated across ancestries. Cross-ancestry fine-mapping, in silico mutagenesis analysis3, and transcriptome-wide and proteome-wide association analyses revealed putative causal genes (such as SH3PXD2A and FURIN) and variants (such as at GRK5 and NOS3). Using a three-pronged approach4, we provide genetic evidence for putative drug effects, highlighting F11, KLKB1, PROC, GP1BA, LAMC2 and VCAM1 as possible targets, with drugs already under investigation for stroke for F11 and PROC. A polygenic score integrating cross-ancestry and ancestry-specific stroke GWASs with vascular-risk factor GWASs (integrative polygenic scores) strongly predicted ischaemic stroke in populations of European, East Asian and African ancestry5. Stroke genetic risk scores were predictive of ischaemic stroke independent of clinical risk factors in 52,600 clinical-trial participants with cardiometabolic disease. Our results provide insights to inform biology, reveal potential drug targets and derive genetic risk prediction tools across ancestries. Stroke is the second leading cause of death worldwide, responsible for prediction of stroke with polygenic scores (PGSs) across ancestries in approximately 12% of total deaths, with an increasing burden particu- the setting of both population-based studies and clinical trials. larly in low-income countries 6. Characterized by a neurological deficit of sudden onset, stroke is predominantly caused by cerebral ischaemia (of which the main aetiological subtypes are large-artery atherosclerotic Genetic discovery from GWASs stroke (LAS), cardioembolic stroke (CES), and small-vessel stroke (SVS)) We performed a fixed-effect inverse-variance weighted (IVW) GWAS and, less often, by intracerebral haemorrhage (ICH). The frequency meta-analysis on 29 population-based cohorts or biobanks with inci- of stroke subtypes differs between ancestry groups as exemplified by dent stroke ascertainment and 25 clinic-based case–control studies, a higher prevalence of SVS and ICH in Asian and African populations comprising up to 110,182 patients who had a stroke and 1,503,898 compared with European populations. Most genetic loci associated control individuals (of whom 45.5% were in longitudinal cohorts or with stroke have been identified in populations of European ancestry. biobanks), nearly doubling the number of cases in previous stroke The largest published GWAS meta-analysis to date (67,162 cases and GWASs (the GIGASTROKE initiative; Supplementary Table 1 and 454,450 control individuals, MEGASTROKE) reported 32 stroke risk Extended Data Fig. 1). Genome-wide genotyping and imputation charac- loci1. To identify new genetic associations and provide insights into teristics are described in Supplementary Table 2. The cohorts included stroke pathogenesis and putative drug targets, we first performed individuals of European (66.7% of the patients who had a stroke), East a cross-ancestry GWAS of 1,614,080 participants, including 110,182 Asian (24.8%), African American (3.7%), South Asian (3.3%) and Hispanic patients who had a stroke, and followed up genome-wide significant (1.4%) ancestry. Analyses were performed for any stroke (AS; compris- signals in an independent dataset of 89,084 patients who had a stroke ing ischaemic stroke, ICH, and stroke of unknown or undetermined and 1,013,843 control individuals. We then characterized the identified type), any ischaemic stroke regardless of subtype (AIS; n = 86,668) stroke risk loci by leveraging expression and protein quantitative trait and ischaemic stroke subtypes (LAS, n = 9,219; CES, n = 12,790; SVS, loci, cross-ancestry fine-mapping and shared genetic variation with n = 13,620). We also conducted separate GWAS analyses of incident other traits. Finally, we used a series of approaches for genomics-driven AS and AIS (n = 32,903 and n = 16,863, respectively) in longitudinal drug discovery for stroke prevention and treatment, and examined the population-based cohort studies. A list of authors and their affiliations appears online. ✉e-mail: Martin.Dichgans@med.uni-muenchen.de; stephanie.debette@u-bordeaux.fr Nature | Vol 611 | 3 November 2022 | 115 Article 1 2 3 4 5 6 7 8 9 10 11 PRDM16 KCNK3* NOA1* FGD5 CASZ1* FOXF2* HDAC9* JCAD THADA FGF5* BNC2 LSP1 EVX1 DEFB1 INPP5B LIPA* TAP1* USP34 DAZL PITX2* CDKN2B-AS1* TMEM41B* COBL WNT2B* SH3PXD2A* FIGN PCDH18* TRIM36 SLC22A7 EBF2 PTCH1 SWAP70* CDK6* IL6R RBM20 3p12 LOC100505841* EDNRA* PIK3CG* SHOC1 OVOL1* PMF1* CENPQ* FBN2 GRK5* NBEAL1* PTK2* THAP5 ATP1B1* FGG* PLSCR5 ARHGAP26 ABO* MMP12* LPA* HTRA1* PRRX1 NOS3* FN1 F11* LAMC1* 12 13 14 15 16 17 18 19 20 CROSS-ANC AS PDE3A* AIS EUR POLR2A CES ANKRD33 FLT1 MESDC1 ZFHX3* EAS LAS HOXC4 DRC3* RMC1 SLC44A2* AFR/SAS SVS LRCH1 FURIN CDH13 PROCR* TSPAN19 DYM LDLR* RPRML ATP2B1 HHIPL1 COL4A2* LINC00924* ZCCHC14* SH2B3* MTMR4* PTPN11 12q24* NCOR2 Fig. 1 | Identifying genetic variants that influence stroke risk. Ideogram Loci are characterized as follows, on the basis of replication results (Methods): showing 89 genome-wide significant stroke-risk loci. The shapes correspond bold with asterisk, high confidence; bold without asterisk, intermediate to ancestry: circles, cross-ancestry (CROSS-ANC); diamonds, Europeans (EUR); confidence; not bold, low confidence; underlined, loci identified in secondary triangles, East Asians (EAS); squares, African Americans (AFR) or South Asians MR-MEGA and MTAG analyses. Black and grey font indicate new and known loci, (SAS). Colours correspond to stroke types: green, AS; red, AIS; light blue, SVS; respectively. The numbers at the top indicate the chromosome. dark blue, CES; purple, LAS. The nearest genes to lead variants are displayed. We tested up to around 7,588,359 single-nucleotide polymorphisms Supplementary Table 5). We also performed cross-ancestry gene-based (SNPs) with a minor allele frequency (MAF) of ≥0.01 for association association tests using VEGAS28 and MAGMA9, which revealed 267 with stroke. The linkage-disequilibrium score intercepts for our gene-wide significant associations (P < 2.63 × 10−6) at 39 loci, of which ancestry-specific GWAS meta-analyses ranged from 0.91 to 1.12, sug- 14 were in 8 new loci that did not reach genome-wide significance in gesting that there was no systematic inflation of association statistics the single-variant analyses (AGAP5/SYNPO2L/SEC24C/CHCHD1, CD96, (Supplementary Table 3). By performing IVW GWAS meta-analyses, we HNRNPA0, MAMSTR, PPM1H, RALGAPA1, USP34 and USP38; Supple- identified variants associated with stroke at genome-wide significance mentary Tables 6 and 7). (P < 5 × 10−8) at 60 loci, of which 33 were new (Fig. 1 and Supplementary Next, we conducted a secondary cross-ancestry GWAS meta-analysis Table 4). Lead variants at all of the new loci were common (MAF ≥ 0.05), using MR-MEGA10, which accounts for the allelic heterogeneity between except for low-frequency intronic variants in THAP5 (MAF = 0.02, in ancestries. We identified three additional genome-wide significant complete association (r2 = 1) with variants in the 5′ UTR of NRCAM) asso- loci for AS (all new), near TSPAN19, and in introns of DAZL and SHOC1, ciated with cross-ancestry incident AS/AIS, and in COBL (MAF = 0.04) all showing high heterogeneity in allelic effects across ancestries (het- associated with AS/AIS in South Asian individuals. Most of the associa- erogeneity P < 0.01; Supplementary Table 8). To further enhance the tions for these 60 loci were with AS (48 loci, 23 new) and AIS (45 loci, statistical power for AIS subtypes, we conducted secondary multitrait 18 new), and one of the AIS loci was associated only with incident AIS analyses of GWASs (MTAG)11 in Europeans and East Asians, including (Supplementary Table 4c). Although AIS subtypes were not available in traits correlated with specific stroke subtypes, namely (1) coronary some population-based cohorts (Supplementary Table 1), genome-wide artery disease (CAD) for LAS, both caused by atheroma; (2) atrial significance was reached for 4 loci for LAS, 8 for CES and 7 for SVS (of fibrillation for CES, as its main underlying cause; and (3) white mat- which 1, 3 and 3 were new, respectively; Supplementary Table 4). Our ter hyperintensity volume (WMH, an MRI-marker of cerebral small results include a large and comprehensive description of stroke genetic vessel disease) for SVS (available in Europeans only). In Europeans, risk variants in each of the five represented ancestries. In cross-ancestry 11 additional loci were associated with LAS (10 new), 3 with SVS (all meta-analyses, 53 loci (51 loci after controlling for ancestry-specific reported in a recent SVS GWAS2) and 5 with CES (all new; Supplementary linkage-disequilibrium score intercepts) reached genome-wide sig- Tables 9–11). Moreover, 18 and 15 additional genome-wide significant nificance (Supplementary Table 4), whereas 42 loci were genome-wide associations were identified (all new) for AS and AIS, respectively, using significant in individual ancestries (35 in Europeans, 6 in East Asians, MTAG with WMH, CAD and atrial fibrillation (Supplementary Tables 12 1 in South Asians and 2 in African Americans; Supplementary Table 4). and 13). In East Asian individuals, one locus was associated with AS Using conditional and joint analysis (GCTA-COJO)7, we confirmed three (FGF5) and one with LAS (HDAC9, new in East Asians) using MTAG. This independent signals at PITX2 and two at SH3PXD2A1 (CES in Europeans; brings the number of identified stroke-risk loci from primary (IVW) 116 | Nature | Vol 611 | 3 November 2022 and secondary (MR-MEGA and MTAG) analyses to 89 in total (61 new), (PIP) of >0.9. We found overlapping credible sets between European of which 69 were associated with AS, 45 with AIS, 15 with LAS, 13 with and East Asian participants at SH3PXD2A (19 overlapping variants), CES and 10 with SVS (of these 44, 33, 11, 8 and 3 were new, respectively; suggesting that there is cross-ancestry-shared genetic architecture Fig. 1 and Supplementary Tables 4, 8 and 9–14). at this locus (Fig. 2b). Two loci had credible sets with a single variant (rs10886430 at GRK5 (PIP = 0.999), associated with GRK5 platelet gene expression and thrombin-induced platelet aggregation13, and rs1549758 Independent follow-up of GWAS signals at NOS3, PIP = 0.995), probably representing strong targets for func- We followed up genome-wide significant stroke-risk loci both inter- tional validation. nally and externally. First, we sought to replicate the 42 stroke-risk Although there were six non-synonymous variants among cred- loci that reached genome-wide significance in individual ancestries ible sets (rs671 (ALDH2), rs8071623 (SEPT4), rs35212307 (WDR12), in at least one other ancestry group among the discovery samples. We rs72932557 (CARF), rs11906160 (MYH7B) and rs2501968 (CENPQ)), successfully replicated, with consistent directionality, 10 of these loci exonic variants for coding RNA within credible sets were few (1.2%). To at P < 1.19 × 10−3 (accounting for the number of loci tested), of which 7 detect putative causal regulatory variants, we conducted an in silico were genome-wide significant in Europeans, 1 in East Asians, and 2 in mutagenesis analysis using MENTR, a machine-learning method to both Europeans and East Asians. An additional 15 loci showed nominal precisely predict transcriptional changes caused by causal variants3. association (P < 0.05) in at least one other ancestry (Supplementary From credible sets, we obtained 78 robust predictions of variant– Table 15). transcript-model sets comprising 13 variants and 19 transcripts (Sup- Second, we gathered an independent dataset of 89,084 individuals plementary Table 20), involving multiple cell types, consistent with the who had a stroke (AS; of which 85,546 AIS; 70.0% European, 15.6% African diversity of mechanisms that underlie stroke aetiology. For example, American, 10.1% East Asian, 4.1% Hispanic and 0.1% South Asian) and the G allele of rs12476527 (5′ UTR of KCNK3) is a risk allele for stroke and 1,013,843 control individuals, mostly from large biobanks, for external was predicted to increase KCNK3 expression in kidney cortex tubule replication (the biobank setting did not allow suitable ischaemic stroke cells, despite no expression quantitative trait loci (eQTL) of this variant subtype analyses). Out of the 60 loci that reached genome-wide sig- being reported in Genotype-Tissue Expression (GTEx, v.8) or eQTLgen nificance in the IVW meta-analyses, 48 loci (80%) replicated at P < 0.05 (2019-12-23). The same G allele has been associated with higher systolic with consistent directionality (Extended Data Fig. 2), of which 31 (52%) blood pressure14. Furthermore, three variants (rs12705390 at PIK3CG, replicated at P < 8.2 × 10−4 (accounting for the number of loci tested) rs2282978 at CDK6 and rs2483262 at PRDM16) were predicted to affect (Supplementary Table 16). When considering both the internal and the expression of a long non-coding RNA and enhancer RNAs, predomi- external follow-up, 52 (87%) of the 60 IVW loci replicated, of which nantly in endothelial cells, as well as other vascular cells and visceral 37 replicated with high confidence, and 15 with intermediate confi- preadipocytes, whereas a promoter variant of SH3PXD2A was predicted dence (Methods, Fig. 1 and Supplementary Table 14). The 8 loci that to modulate its expression in macrophages. did not replicate were labelled as low confidence (Methods and Sup- plementary Table 14). Four of these were ethnic specific and three were low-frequency variants that were monomorphic in some ancestries and Characterizing stroke-associated loci were therefore probably underpowered for replication. VEGAS2Pathway 15 analysis revealed significant enrichment Within the secondary analyses, none of the three MR-MEGA loci (P < 5.01 × 10−6) of stroke-risk loci in pathways involved in (1) carboxy- replicated, although one was borderline significant (Supplementary lation of amino-terminal glutamate residues required for the activation Table 16). Of the 26 MTAG loci, 18 (69%) replicated with AS or AIS at of proteins involved in blood clot formation and regulation; (2) negative P < 0.05, of which 9 (35%) replicated with high confidence (P < 1.7 × 10−3, regulation of coagulation; and (3) angiopoietin receptor Tie2-mediated accounting for 29 secondary loci tested; Supplementary Table 16). Of signalling, involved in angiogenesis (Supplementary Table 21). the eight MTAG loci that did not replicate, seven showed a consist- We examined shared genetic variation with 12 (in Europeans) and 10 ent directionality and four were subtype specific and were therefore (in East Asians) vascular risk factors and disease traits (Methods and underpowered to detect associations with AS or AIS. Supplementary Methods). In Europeans, the lead variants for stroke at 57 of the 89 primary and secondary risk loci (64.0%) were associated (P < 5 × 10−8) with at least one vascular trait, most frequently blood pres- Cross-ancestry effects and fine-mapping sure (33 loci, 37.1%; Extended Data Fig. 4 and Supplementary Table 22). For the 60 loci associated with stroke risk derived from the IVW After correction for multiple testing (Methods; P < 4.17 × 10−3), all of the meta-analyses, we first demonstrated the added value in terms of locus vascular-risk traits except for low-density lipoprotein (LDL)-cholesterol discovery of including non-European samples, showing a clear gain in showed significant genetic correlation (rg) with at least one stroke power beyond sample size increase, compared with the incremental type, the strongest correlations being for CAD and LAS (rg = 0.73), atrial addition of European ancestry samples (Extended Data Fig. 3). We next fibrillation and CES (rg = 0.63), and systolic blood pressure (SBP) with compared the per-allele effect size across the three ancestries with the all stroke types (rg ranging from 0.21 for CES to 0.49 for LAS and SVS; largest sample size (European, East Asian, African American). Corre- Extended Data Fig. 5 and Supplementary Table 23). Using two-sample lations of per-allele effect sizes of index variants varied from r = 0.55 Mendelian randomization (MR), we found evidence for a possible causal (European with African American) to r = 0.66 (European with East Asian) association for every vascular-risk trait except for triglycerides with at and r = 0.74 (East Asian with African American; Fig. 2a). least one stroke type (P < 4.17 × 10−3), with some subtype-specific asso- To identify putative causal variants at stroke-risk loci identified ciation patterns. Genetic liability to WMH was associated with increased through IVW meta-analyses, we performed multiple-causal-variant risk of SVS but not other stroke subtypes, whereas genetic liability to fine-mapping using SuSiE12, separately in European and East Asian venous thromboembolism was associated with AS, AIS, CES and LAS, participants (Methods). Across stroke types, we identified 110 and 16 but not SVS (Extended Data Fig. 5 and Supplementary Table 24). Owing 95% credible set–trait pairs in European and East Asian participants, to a limited overlap between the European GIGASTROKE sample and respectively, each of which having a 95% posterior probability of con- cohorts included in GWASs for the exposure traits, we ran sensitivity taining a causal variant, with multiple credible sets identified at 6 (in analyses weighting our genetic instruments on the basis of a sub-sample Europeans) and 1 (in East Asians) stroke-risk loci (Supplementary of the UK Biobank, excluding cases included in GIGASTROKE16. The Tables 17–19). Within the credible sets identified in European partici- notable consistency of these with the main analyses confirmed their pants, 17 variants were found to have a posterior inclusion probability robustness against weak instrument bias (Supplementary Table 25). Nature | Vol 611 | 3 November 2022 | 117 Article a 0.4 0.4 0.4 r = 0.66 r = 0.55 r = 0.74 Both 0.2 0.2 0.2 CDK6 KCNK3 EUR only ANKRD33 LRCH1 PRDM16 KCNK3 THAP5 PROCR EAS only Effect EAS Effect AFR Effect AFR BNC2 GRK5 PRDM16 0 0 0 AFR only PITX2 FOXF2 SLC44A2 LRCH1 Cross- TAP1 TAP1 ancestry –0.2 –0.2 EVX1 –0.2 only COBL NS –0.4 –0.4 –0.4 –0.2 –0.1 0 0.1 0.2 –0.2 –0.1 0 0.1 0.2 –0.2 –0.1 0 0.1 0.2 Effect EUR Effect EUR Effect EAS b R2 with cross-ancestry lead variant 10.0 Cross-ancestry lead variant 7.5 –log10[P] 5.0 1.0 EUR 2.5 0.8 0 0.3 0.6 rs55983834 Overlapping CS 0.2 0.4 EUR PIP 0.1 0.2 0 0 10.0 –log10[P] EAS 5.0 0 0.10 rs4918058 EAS PIP 0.05 0 6.0 4.0 –log10[P] HIS 2.0 0 6.0 4.0 –log10[P] AFR 2.0 0 6.0 4.0 –log10[P] SAS 2.0 0 STN1 SLK SH3PXD2A-AS1 MIR936 COL17A1 SH3PXD2A 105.45 105.5 105.55 105.6 105.65 105.7 105.75 105.8 105.85 Chromosome 10 position (Mb) Fig. 2 | Effect-size comparison across ancestry groups of lead variants purple, European only (±cross-ancestry); green, East Asian only identified in stroke GWASs and cross-ancestry fine-mapping. a, Plots showing (±cross-ancestry); yellow, African American only (±cross-ancestry); blue, the Pearson’s correlation coefficient (r) between the effect sizes (β) of the 60 both ancestries (±cross-ancestry); red, cross-ancestry only; grey, not stroke-risk alleles on AS significant after multiple-testing correction (P < 0.017) genome-wide significant in two plotted ancestries and in cross-ancestry. in Europeans and East Asians (left; r (95% CI) = 0.66 (0.47–0.79), P = 1 × 10 −7); b, Locus plots of variants at SH3PXD2A in five ancestries. Fine-mapped variants Europeans and African Americans (middle; r (95% CI) = 0.55 (0.33–0.71), are shown only in European and East Asian individuals (insufficient power P = 2 × 10 −5); and East Asians and African Americans (right; r (95% CI) = 0.74 for other ancestries). Variants are coloured on the basis of their linkage (0.58–0.85), P = 8 × 10 −10). n = 60 independent stroke-risk variants from the IVW disequilibrium with the cross-ancestry lead variant (rs4918058), shown by the meta-analyses were used to compute Pearson’s correlation coefficients (r) of purple diamonds. In the fine-mapping plots, variants in the SuSiE 95% credible the effect sizes between ancestries. The nearest gene is reported for SNPs sets (CS) are shown. Shared variants between credible sets of European and showing a difference in effect size (β, absolute value) of >0.05 between a pair East Asian participants are indicated by black circles. The red vertical lines of ancestries. The dots represent the effect-size (β) estimates and the bars represent the position of the lead variants in European (rs55983834) and East represent the 95% CI of the estimates. Two-sided P values of the deviation of Asian (rs4918058) participants. The grey dashed horizontal lines represent Pearson’s correlation coefficient from zero are reported. Colour corresponds P = 5 × 10 −8. The linkage disequilibrium of each ancestry was derived from the to genome-wide significant association (P < 5 × 10 −8) in individual ancestries: 1000 Genomes Project. We confirmed directionality using the Steiger test (Supplementary AIS and SVS; CAD with AS, AIS and LAS; and atrial fibrillation with CES Table 24) and ruled out reverse causation with reverse MR (Supple- (Extended Data Fig. 6 and Supplementary Tables 23 and 24). Notably, mentary Table 26). In East Asian individuals, SBP, diastolic blood MR analyses performed with binary exposures should be interpreted pressure (DBP), body mass index (BMI) and atrial fibrillation showed with caution owing to the potential violations of the exclusion restric- significant genetic correlation with AS (rg = 0.45, 0.39, 0.24 and 0.32 tion assumption16. versus rg = 0.36, 0.21, 0.22 and 0.44 in Europeans) and AIS (except for Next, to generate hypotheses of target genes and directions of effect, BMI), with evidence for a causal association of SBP and DBP with AS, we conducted transcriptome-wide association studies (TWAS) using 118 | Nature | Vol 611 | 3 November 2022 Methods Drug-target genes Example candidate drugs (Anatomical Therapeutic Chemical Classification System B drugs, for FGA Alteplase AS, AIS and CES). This encompasses the previously described PDE3A and FGA genes1, which encode targets for cilostazol (antiplatelet agent) and PDE3A Cilostazol alteplase (thrombolytic drug acting through plasminogen23), respec- tively, as well as F11, KLKB1, F2, TFPI and MUT, which encode targets MUT Cyanocobalamin, hydroxocobalamin for conestat alfa, ecallantide (both used for hereditary angioedema), Gene prioritization and enrichment in ATC code F2 Lepirudin lepirudin, dalteparin (both used to treat recurrent thromboembolism) (GREP) and vitamin B12, respectively (Supplementary Table 31). Notably, the TFPI Dalteparin results for AS are probably driven by AIS (the vast majority of AS in the current study) and cannot be extrapolated to ICH. Second, we used F11 Abelacimab, conestat alfa Trans-Phar24 to test the negative correlations between genetically deter- mined case–control gene expression associated with stroke (TWAS KLKB1 Isolated human kallikrein using all GTEx v.7 tissues17) and compound-regulated gene expression profiles. At FDR < 0.10, we observed significant negative correlations PROC Ancrod for BRD.A22514244 (for SVS; drug target unknown) and GR.32191 (for MR and colocalization CES; Supplementary Table 32). GR-32191 is a thromboxane A2 receptor VCAM1 Symbiopolyol, probucol antagonist that has been proposed as an alternative antiplatelet therapy for stroke prevention25, and further drugs of this class are under devel- GP1BA Anfibatide, TGX-6b4 opment26. Note that one of those drugs, terutroban, was evaluated in a LAMC2 BST-1005 phase III study but did not show non-inferiority against aspirin27. Third, we used protein quantitative trait loci (pQTL) for 218 drug-target pro- Negative correlation tests GR-32191 teins as instruments for MR and found evidence for causal associations between genetically determined and compound-regulated of 9 plasma proteins with stroke risk (4 cis-pQTL and 6 trans-pQTL), of gene expression (Trans-Phar) BRD-A22514244 which 7 were supported by colocalization analyses, with no evidence for reverse causation using the Steiger test (PROC, VCAM1, F11, KLKB1, AIS and AS SVS CES MMP12, GP1BA and LAMC2; Supplementary Table 33). All of these rep- Fig. 3 | Genomics-driven drug discovery. Overlap enrichment analysis using licated (at FDR < 0.05) with consistent directionality using at least one GREP22 (top). Middle, integrating MR results using cis- and trans-pQTLs as independent plasma pQTL resource and cerebrospinal fluid pQTL instrumental variables with data from drug databases. Bottom, negative for PROC and KLKB1, with evidence for colocalization for PROC, F11, correlation tests between compound-regulated gene expression profiles and KLKB1 and MMP12, but not for GP1BA (for which both concordant and genetically determined case–control gene expression profiles using discordant directionality was observed) and LAMC2 (pQTL available in Trans-Phar. one replication dataset only; FDR = 0.08). Using public drug databases, we curated drugs targeting those proteins in a direction compatible with a beneficial therapeutic effect against stroke based on MR esti- TWAS-Fusion and eQTL based on RNA-sequencing (RNA-seq) analyses mates and identified such drugs for VCAM1, F11, KLKB1, GP1BA, LAMC2 in different tissues17–20. We identified 27 genes of which the geneti- (inhibitors) and PROC (activators; Supplementary Table 34). Drugs cally regulated expression is associated with stroke and its subtypes targeting F11 (NCT04755283, NCT04304508, NCT03766581) and PROC at the transcriptome-wide level and colocalized in at least one tissue (NCT02222714) are currently under investigation for stroke, and our (10 genes in arteries and heart; 6 genes in brain tissue; 17 genes across results provide genetic support for this. Notably, F11 and KLKB1 are tissues). Of these genes, 18 overlapped with 11 genome-wide significant adjacent genes with a long-range linkage-disequilibrium pattern and stroke-risk loci (Extended Data Fig. 7 and Supplementary Table 27). For complex co-regulation28, as illustrated here by the presence of a shared several genes of which bulk tissue expression levels showed evidence trans-pQTL in KNG1 (Supplementary Table 33). Additional studies are for association with stroke, human single-nucleus sequencing data of needed to disentangle causal associations and the most appropri- brain cells in the dorsolateral prefrontal cortex (DLPFC) showed distinct ate drug target in this region29,30. Next, for the five genes targeted by cell-specific gene expression patterns suggesting that multiple genes inhibitors, VCAM1, F11, KLKB1, GP1BA and LAMC2, we examined the could be involved through different cell types21 (Extended Data Fig. 8). associations of rare deleterious variants (MAF < 0.01) with stroke and Overall, we observed a significant enrichment mostly in brain vascular stroke-related traits, applying gene-based burden tests to whole-exome endothelial cells and astrocytes, possibly reflecting the importance sequencing data from >450,000 UK Biobank participants to support of both vascular pathology and brain response to the vascular insult potential therapeutic targets for inhibitors31. We observed one sig- in modulating stroke susceptibility (Extended Data Fig. 8 and Sup- nificant protective association of rare deleterious variants in F11 with plementary Tables 28 and 29). Furthermore, using proteome-wide venous thromboembolism (odds ratio (OR) = 0.471, P = 2.46 × 10−4), association studies (PWAS) in DLPFC brain tissue, we found evidence in a direction concordant with that of MR estimates (Supplementary for the association of ICA1L with AS and AIS through its cis-regulated Table 35). To further validate the candidate drugs and estimate their protein abundance, with colocalization evidence (Extended Data Fig. 8 potential side effects, we investigated whether the drug-target genes and Supplementary Table 30). In both TWAS and PWAS, lower ICA1L were associated with stroke-related phenotypes using a phenome-wide transcript or protein abundance in the DLPFC was associated with a association study (PheWAS) approach. We conducted PheWAS in the higher risk of stroke. Estonian Biobank (EstBB) for pQTL variants for the PROC, VCAM1, F11, KLKB1, GP1BA and LAMC2 genes. A cis-pQTL for F11, rs2289252, was associated with higher risk of venous thromboembolic disorders Genomics-driven drug discovery (P < 3.45 × 10−6), as previously described32, and showed suggestive asso- We used a three-pronged approach for genomics-driven discovery ciation (P = 3.44 × 10−3) with cerebral artery occlusion with cerebral of drugs for the prevention or treatment of stroke4 (Methods and infarction (Phecode 433.21; Extended Data Fig. 9 and Supplementary Fig. 3). First, using GREP22, we observed significant enrichment of Table 36). By contrast, we observed no significant association with stroke-associated genes (MAGMA9 or VEGAS28 false-discovery rates non-stroke-related phenotypes, suggesting the safety of targeting F11. (FDR) < 0.05) in drug-target genes for blood and blood-forming organs Similar profiles were observed in the UK Biobank (https://pheweb.org/ Nature | Vol 611 | 3 November 2022 | 119 Article UKB-SAIGE/variant/4-187207381-C-T) and FinnGen (https://r7.finngen. was 1.33 (95% CI = 1.26–1.40, P = 9.9 × 10−26) for the iPGS model. The fi/variant/4-186286227-C-T), with no significant associations with other MEGASTROKE- and GIGASTROKE-based iPGS models for Europeans disorders and no overlap of subthreshold signals with side-effects achieved a lower AUC improvement (∆AUC = 0.007 and 0.009, respec- reported in clinical trials33. We further confirmed the association of tively) than the GIGASTROKE-based iPGS model for East Asians. While rs2289252 with venous thromboembolic disorders and that it has this suggests that the transferability of iPGS models from Europeans to no association with other non-stroke-related phenotypes using the East Asians might be limited (Supplementary Table 45), it does indicate Phenoscanner database (Supplementary Table 37). that an ancestry-specific stroke iPGS approach yields similar improve- Overall, combining evidence from genomics-driven drug discovery ment in predictive ability relative to their base models. approaches, characterization of stroke-risk loci (missense variants, Participants in the top 1% of the iPGS showed 1.9-fold higher odds of TWAS, PWAS, colocalization, pathway enrichment, MR with pQTL, AIS (OR = 1.90, 95% CI = 1.20–2.91, P = 0.004) compared with the middle MENTR and PoPS34), and previous knowledge from monogenic disease 10% (Fig. 4b and Supplementary Table 46). We further confirmed the models and experimental data, we found evidence for the potential GIGASTROKE-based East Asian iPGS model trained on the BBJ in 1,399 functional implication of 56 genes that should be prioritized for further cases of prevalent AIS and 86,283 controls from the Taiwan Biobank functional follow-up, with evidence from multiple approaches for 20 (TWB): OR per s.d. = 1.18 (95% CI = 1.12–1.25, P = 1.1 × 10−9), with a ∆AUC genes (Supplementary Table 38). of 0.003 (Supplementary Table 47). Notably, iPGS models derived from cross-ancestry stroke GWASs had a higher predictive ability compared with iPGS models derived from Integrative polygenic risk prediction ancestry-specific stroke GWASs both in Europeans and East Asians We investigated the risk prediction potential of stroke GWASs, alone (Supplementary Table 48). and in combination with vascular-risk-trait GWASs, first in Europeans Next, we evaluated the predictive ability of the European-derived and East Asians, using ancestry-specific PGSs. PGSs were based on GIGASTROKE-based iPGS model in African American and indig- ancestry-specific and cross-ancestry GWAS summary statistics. We first enous African (Nigerian and Ghanaian) datasets. In 107,343 African derived single PGS (sPGS) models from single stroke GWAS summary American MVP participants, of whom 2,227 developed an AIS, the data (Supplementary Table 39). We then constructed integrative PGS GIGASTROKE-based iPGS model showed a significant association (iPGS) models, which combined multiple GWAS summary data of dif- with AIS incidence (HR per 1 s.d. = 1.11, 95% CI = 1.06–1.17, P = 1.8 × 10−5, ferent traits into a PGS using elastic-net logistic regression5 (Extended ∆C-index = 0.003; Supplementary Table 49), although weaker than Data Fig. 10). The iPGS analysis used two datasets for each ancestry for in European MVP participants (Supplementary Table 43). The par- model training and evaluation, respectively. The participants in the ticipants in the top 1% of the iPGS showed 1.5-fold higher odds of AIS training and evaluation datasets did not overlap and were not included (HR = 1.53, 95% CI, 1.04–2.25, P = 0.03) compared with participants in in the input GWAS summary data. the middle 10% (Fig. 4c and Supplementary Table 50). In 1,691 cases For Europeans, we constructed the iPGS model using 1,003 prevalent and 1,743 control participants from the indigenous African (Nigerian AIS cases and 8,997 controls, followed by evaluation of the model using and Ghanaian) SIREN case–control study, the GIGASTROKE-based 1,128 incident AIS cases among 102,099 participants, all from the EstBB. iPGS also showed a significant association with the odds of AIS (OR per The improvement in predictive ability (∆C-index) was assessed over 1 s.d. = 1.09, 95% CI = 1.02–1.17, P = 0.010, ∆AUC = 0.007; Supplemen- a base model including age, sex and the top 5 principal components tary Table 51). The GIGASTROKE-based iPGS model showed a stronger (PCs) for population stratification. The iPGS model for Europeans association with AIS and a larger improvement in predictive ability incorporated 10 GIGASTROKE GWAS analyses (all stroke types, using compared with the MEGASTROKE-based iPGS model in both MVP and the European and cross-ancestry analysis) and 12 vascular-risk-trait SIREN (Supplementary Tables 49 and 51). GWAS analyses (Extended Data Fig. 10 and Supplementary Table 40). The iPGS model achieved a ∆C-index of 0.027 (Supplementary Table 41), 93% higher than that for a previously constructed iPGS Risk prediction in clinical trials model for Europeans, derived from 5 MEGASTROKE GWAS analyses Following up on previous work1,35, we further examined whether a and similar vascular-risk-trait GWASs (∆C-index = 0.014)5. The age-, genetic risk score (GRS) based on genome-wide significant risk loci from sex- and top 5 PC-adjusted hazard ratio (HR) per s.d. of the iPGS was the cross-ancestry IVW AS meta-analyses could identify individuals who 1.26 (95% confidence interval (CI) = 1.19–1.34, P = 2.0 × 10−15) for the are at higher risk of AIS after accounting for established risk factors in GIGASTROKE-based iPGS model compared to 1.19 (95% CI = 1.12–1.26, five clinical trials across the spectrum of cardiometabolic disease35. P = 4.2 × 10−9) for the MEGASTROKE-based iPGS model. Compared with The primary analysis was conducted in 51,288 European participants participants in the middle 10% (45–55%) of the GIGASTROKE-based of whom 960 developed an incident ischaemic stroke (AIS) over a 3 year iPGS model, those in the top 1% showed a >2.5-fold higher hazard of AIS follow-up. In a Cox model adjusted for age, sex and vascular risk factors (HR = 2.56, 95% CI = 1.59–4.10, P = 9.6 × 10−5; Fig. 4a and Supplementary (Methods), a higher GIGASTROKE GRS was significantly associated with Table 42). We further confirmed the GIGASTROKE-based European increased risk of AIS in Europeans (adjusted HR = 1.17, 95% CI = 1.09–1.24 iPGS model trained on the EstBB in 403,489 European-ancestry par- per s.d. increase, P = 2 × 10−6; Supplementary Table 52). This associa- ticipants of the Million Veteran Program (MVP) study, of whom 8,392 tion was substantially stronger than the association with the earlier developed an AIS: HR per s.d. = 1.19 (95% CI = 1.16–1.21, P = 6.94 × 10−52), MEGASTROKE GRS based on 32 genome-wide significant stroke-risk loci with a ∆C-index of 0.010 (Supplementary Table 43). (HR = 1.07, 95% CI = 1.00–1.14, P = 0.036)1,35. Compared with patients in For East Asians, we derived the iPGS model using 577 cases of preva- the lowest GIGASTROKE GRS tertile, patients in the top GRS tertile had lent AIS and 9,232 control individuals, and evaluated the model using an adjusted HR of 1.35 (95% CI = 1.16–1.58) for developing AIS, whereas 1,470 cases of prevalent AIS and 40,459 control individuals from those in the middle tertile had an adjusted HR of 1.13 (95% CI = 0.96–1.33, Biobank Japan (BBJ). A base model including age, sex and the top 5 Ptrend = 1.4 × 10−4; Fig. 4e). The performance of the GRS was stronger in PCs showed an area under the curve (AUC) of 0.634. The iPGS model individuals who had not previously had a stroke (n = 44,095; adjusted was constructed by integrating 10 GIGASTROKE GWAS analyses HR of the top versus lowest tertile = 1.37, 95% CI = 1.14–1.65) compared and 12 vascular-risk-trait GWAS analyses (Extended Data Fig. 10 and with in those who previously had a stroke (n = 7,193; adjusted HR = 1.15, Supplementary Table 44). The iPGS model for East Asians showed 95% CI = 0.87–1.54). Similar associations were observed when using an improvement in AUC (∆AUC) of 0.019 (Supplementary Table 45). effect estimates from stroke GWAS meta-analyses in Europeans or for The age-, sex- and top 5 PC-adjusted odds ratio (OR) per s.d. of PGS AIS (Supplementary Table 52). In secondary analyses, we examined 120 | Nature | Vol 611 | 3 November 2022 a 45–55%, reference group d 45–55%, reference group 55–100%, HR = 1.29 55–100%, HR = 1.28 80–100%, HR = 1.46 80–100%, HR = 1.38 0.4 0.4 90–100%, HR = 1.57 90–100%, HR = 1.46 0.3 95–100%, HR = 1.74 0.3 95–100%, HR = 1.75 97.5–100%, HR = 2.11 97.5–100%, HR = 1.82 Density Density 0.2 99–100%, HR = 2.56 0.2 99–100%, HR = 2.78 99.5–100%, HR = 2.72 99.5–100%, HR = 3.38 0.1 99.9–100%, HR = 3.17 0.1 99.9–100%, HR = 5.38 0 0 –5.0 –2.5 0 2.5 5.0 –5.0 –2.5 0 2.5 5.0 iPGS (z-score) iPGS (z-score) b 45–55%, reference group e 55–100%, OR = 1.41 3.0 80–100%, OR = 1.64 0.4 90–100%, OR = 1.77 2.5 cumulative event rate (%) 0.3 95–100%, OR = 1.94 97.5–100%, OR = 2.10 Density Ischaemic stroke 2.0 0.2 99–100%, OR = 1.90 99.5–100%, OR = 2.15 0.1 1.5 99.9–100%, OR = 3.40 0 1.0 Low genetic risk –5.0 –2.5 0 2.5 5.0 iPGS (z-score) Int. genetic risk 0.5 High genetic risk c 45–55%, reference group Ptrend < 0.001 55–100%, HR = 1.11 80–100%, HR = 1.14 0 0.4 90–100%, HR = 1.28 0 360 720 1,080 0.3 95–100%, HR = 1.36 Time (days) Density 97.5–100%, HR = 1.38 0.2 99–100%, HR = 1.53 Number at risk 99.5–100%, HR = 1.69 0.1 99.9–100%, HR = 2.18 Low genetic risk 17,096 16,842 16,634 16,256 14,126 8,800 3,707 0 Int. genetic risk 17,096 16,807 16,584 16,193 14,041 8,684 3,558 –5.0 –2.5 0 2.5 5.0 High genetic risk 17,096 16,828 16,574 16,212 13,961 8,649 3,572 iPGS (z-score) Fig. 4 | Risk prediction in a population and trial setting. a–d, The association regression models for East Asian individuals with adjustments for age, sex and of iPGS with ischaemic stroke (AIS) in European (Estonian Biobank) (a), East the top five genetic principal components. e, Kaplan–Meier event rates for Asian (BioBank Japan) (b), African American (Million Veteran Program) (c) and ischaemic stroke in European participants in five clinical trials (Methods) by European participants in clinical trials (d). Compared with the middle decile tertile of GRS at 3 years (the GRS uses effect estimates of the cross-ancestry AS (45–55%) of the population as a reference group, the risk of high-iPGS groups GWAS as weights) showing higher GRS increases risk of ischaemic stroke with varying percentile thresholds was estimated using a Cox proportional (P trend = 1.4 × 10 −4). The two-sided P trendvalue was computed using Cox hazards model for European and African American individuals and logistic regression. Int., intermediate. the association of the GIGASTROKE cross-ancestry AS GRS with showed differences in effect size, for example, at PROCR, TAP1 or incident AIS in the much smaller East Asian sample (1,312 partici- BNCZ-CNTLN (MAF ≤ 0.05 in East Asians), or at GRK5, FOXF2 or COBL pants of whom 27 developed an incident AIS over a 3 year follow-up), (MAF ≤ 0.05 in African Americans). Ancestry-specific meta-analyses and found consistent associations (adjusted HR = 1.49, 95% CI = in smaller non-European populations detected fewer loci than in 1.00–2.21 per s.d. increase, P = 0.048; Supplementary Table 52), whereas Europeans that were nevertheless biologically plausible, for example, the MEGASTROKE GRS was not associated with incident AIS in East 3p12 and PTCH1 for SVS in African Americans. Rare variants at 3p12 Asians (adjusted HR = 0.82, 95% CI = 0.55–1.23, P = 0.34). Finally, in were recently shown to be associated with WMH volume36, whereas European trial participants (there were too few East Asian individuals common variants at PTCH1 were associated with functional outcome for this analysis), the GIGASTROKE-based iPGS was also significantly after ischaemic stroke (in European individuals)37. New association associated with increased AIS incidence (HR per 1 s.d. increase = 1.19, signals from cross-ancestry GWASs included, for example, variants at 95% CI = 1.11–1.27, P = 3.2 × 10−7, ∆C-index = 0.008), performing better PROCR, GRK5 and F11 (thrombosis), LPA and ATP2B1 (lipid metabolism, than the MEGASTROKE-based iPGS (Supplementary Table 53). Com- hypertension and atherosclerosis), SWAP70 (membrane ruffling) and pared with the middle 10% of the participants, those in the top 1% had a LAMC1 (cerebrovascular matrisome). 2.8-fold higher hazard of AIS (HR = 2.78, 95% CI = 1.67–4.61, P = 7.9 × 10−5) Extensive bioinformatics analyses highlight genes for prioritization (Fig. 4d and Supplementary Table 54). in functional follow-up studies (Supplementary Table 38). For example, a promoter variant of SH3PXD2A, which encodes an adaptor protein that is involved in extracellular matrix degradation through invadopodia Discussion and podosome formation, was predicted to modulate its expression in Our GWAS meta-analyses, including 110,182 patients who had a stroke macrophages38. FURIN expression levels across tissues were associated and 1,503,898 control participants from five different ancestries (33% of with an increased stroke risk. FURIN has previously been implicated in patients who had a stroke were non-European), identified 89 (61 new) CAD39 as well as in atherosclerotic lesion progression in mice40. It also risk loci for stroke and stroke subtypes (60 through primary IVW and has a key role in SARS-CoV-2 infectivity41, and patients with COVID-19 29 through secondary MR-MEGA and MTAG analyses). We observed are at increased risk of AIS, especially LAS42; the FURIN locus was pre- substantial shared susceptibility to stroke across ancestries, with a dominantly associated with LAS in our data (Supplementary Table 55). strong correlation of effect sizes. On the basis of internal cross-ancestry Our results provide genetic evidence for putative drug effects using validation and independent follow-up in 89,084 cases of stroke (30% three independent approaches, with converging results from two meth- non-European) and 1,013,843 control individuals, mostly from large ods (gene enrichment analysis and pQTL-based MR) for drugs targeting biobanks with information on AS and AIS only, the level of confidence F11 and KLKB1. F11 and F11a inhibitors (such as abelacimab, BAY 2433334 of these loci was intermediate or high for 87% of primary stroke-risk and BMS-986177) are currently being examined in phase 2 trials for pri- loci and 60% of secondary loci. Effect estimates for variants that were mary or secondary stroke prevention (NCT04755283, NCT04304508, common across ancestries were typically similar, whereas, expectedly, NCT03766581). pQTL-based MR suggested PROC as a potential drug variants that were rare or low frequency in one or more populations target for stroke. A recombinant variant of human activated protein C Nature | Vol 611 | 3 November 2022 | 121 Article (encoded by PROC) was found to be safe for the treatment of acute In conclusion, our genomic findings derived from >200,000 patients ischaemic stroke after thrombolysis, mechanical thrombectomy or who had a stroke worldwide provide critical insights to inform future both in phase 1 and 2 trials (3K3A-APC, NCT02222714)43,44, and is poised biological research on stroke pathogenesis, highlight potential drug for an upcoming phase 3 trial. 3K3A-APC is proposed as a neuropro- targets for intervention and provide tools for genetic risk prediction tectant, with evidence for the protection of white matter tracts and across ancestries. oligodendrocytes against ischaemic injury in mice45. Weaker evidence was found for GP1BA, VCAM1 and LAMC2 as potential drug targets for stroke, with evidence for colocalization in only one pQTL dataset. Online content Anfibatide, a GPIbα antagonist, reduced blood–brain barrier disruption Any methods, additional references, Nature Research reporting sum- after ischaemic stroke in mice46 and is being tested as an antiplate- maries, source data, extended data, supplementary information, let drug in myocardial infarction (NCT01585259). Although specific acknowledgements, peer review information; details of author con- VCAM1 inhibitors are not available, probucol—a lipid lowering drug tributions and competing interests; and statements of data and code with pleiotropic effects including VCAM1 inhibition—was tested for availability are available at https://doi.org/10.1038/s41586-022-05165-3. secondary prevention against atherosclerotic events in patients with CAD (PROSPECTIVE, UMIN000003307)47. 1. Malik, R. et al. Multiancestry genome-wide association study of 520,000 subjects We investigated stroke PGSs across ancestries. PGSs integrating identifies 32 loci associated with stroke and stroke subtypes. Nat. Genet. 50, 524–537 cross-ancestry and ancestry-specific stroke GWASs with vascular- (2018). risk-factor GWASs (iPGS) analyses showed strong prediction of ischae- 2. Traylor, M. et al. Genetic basis of lacunar stroke: a pooled analysis of individual patient data and genome-wide association studies. Lancet Neurol. https://doi.org/10.1016/s1474- mic stroke risk in Europeans and, importantly, in East Asians, in whom 4422(21)00031-4 (2021). stroke incidence is highest6. These results were confirmed in several 3. Koido, M. et al. Predicting cell-type-specific non-coding RNA transcription from genome independent datasets. The iPGS performed better than stroke PGS sequence. Preprint at bioRxiv https://doi.org/10.1101/2020.03.29.011205 (2020). 4. Namba, S. et al. A practical guideline of genomics-driven drug discovery in the era alone and better than the previous best iPGS models in Europeans5. of global biobank meta-analysis. Preprint at medRxiv https://doi.org/10.1101/ The transferability of European-specific iPGS models to East Asians 2021.12.03.21267280 (2021). was limited. While there were not enough African participants to gen- 5. Abraham, G. et al. Genomic risk score offers predictive performance comparable to clinical risk factors for ischaemic stroke. Nat. Commun. 10, 5819 (2019). erate an African-specific stroke PGS, the European iPGS showed a sig- 6. GBD 2019 Stroke Collaborators. Global, regional, and national burden of stroke and its nificant association with AIS in both African American and indigenous risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study African participants, although expectedly weaker than in European 2019. Lancet Neurol. 20, 795–820 (2021). 7. Yang, J. et al. Conditional and joint multiple-SNP analysis of GWAS summary statistics participants. Individuals in the top 1% of the PGS distribution had a 2- to identifies additional variants influencing complex traits. Nat. Genet. 44, 369–375 (2012). 2.5-fold risk of ischaemic stroke in East Asian and European participants 8. Mishra, A. & Macgregor, S. VEGAS2: software for more flexible gene-based testing. Twin compared with those in the middle 10%, whereas this risk was 1.5-fold Res. Hum. Genet. 18, 86–91 (2015). 9. de Leeuw, C. A., Mooij, J. M., Heskes, T. & Posthuma, D. MAGMA: generalized gene-set in African American participants. Although caution is warranted when analysis of GWAS data. PLoS Comput. Biol. 11, e1004219 (2015). interpreting risk estimates owing to the wide CIs, these results suggest 10. Magi, R. et al. Trans-ethnic meta-regression of genome-wide association studies that GIGASTROKE-based iPGS models may be useful to stratify individu- accounting for ancestry increases power for discovery and improves fine-mapping resolution. Hum. Mol. Genet. 26, 3639–3650 (2017). als exposed to genetically high risk of ischaemic stroke, especially in 11. Turley, P. et al. Multi-trait analysis of genome-wide association summary statistics using Europeans and East Asians. Our results highlight the importance of MTAG. Nat. Genet. 50, 229–237 (2018). ancestry-specific and cross-ancestry genomic studies for the transfer- 12. Wang, G., Sarkar, A., Carbonetto, P. & Stephens, M. A simple new approach to variable selection in regression, with application to genetic fine mapping. J. R. Stat. Soc. B 82, ability of genomic risk prediction across populations, and the urgent 1273–1300 (2020). need to substantially increase participant diversity in genomic studies, 13. Rodriguez, B. A. T. et al. A platelet function modulator of thrombin activation is causally linked to cardiovascular disease and affects PAR4 receptor signaling. Am. J. Hum. Genet. especially from the most under-represented regions such as Africa, to 107, 211–221 (2020). avoid exacerbation of health disparities in the era of precision medicine 14. Giri, A. et al. Trans-ethnic association study of blood pressure determinants in over and precision public health48. 750,000 individuals. Nat. Genet. 51, 51–62 (2019). 15. Mishra, A. & MacGregor, S. A novel approach for pathway analysis of GWAS data Finally, leveraging data from 5 clinical trials in 52,600 patients with highlights role of BMP signaling and muscle cell differentiation in colorectal cancer cardiometabolic disease, we showed that a cross-ancestry GRS pre- susceptibility. Twin Res. Hum. Genet. 20, 1–9 (2017). dicted ischaemic stroke, independently of clinical risk factors, and 16. Burgess, S., Davies, N. M. & Thompson, S. G. Bias due to participant overlap in two-sample Mendelian randomization. Genet. Epidemiol. 40, 597–608 (2016). outperforming previous genetic risk evaluation35. Notably, although 17. GTEx Consortium. Genetic effects on gene expression across human tissues. Nature 550, the trials included predominantly European participants, consistent 204–213 (2017). results were observed in East Asian participants. We further confirmed 18. Fromer, M. et al. Gene expression elucidates functional impact of polygenic risk for schizophrenia. Nat. Neurosci. 19, 1442–1453 (2016). the GIGASTROKE iPGS in these clinical trials. 19. Gusev, A. et al. Integrative approaches for large-scale transcriptome-wide association Our study includes a considerable contribution of non-European studies. Nat. Genet. 48, 245–252 (2016). stroke genetics resources (n = 61,528/616,014 cases/controls for the 20. Wright, F. A. et al. Heritability and genomics of gene expression in peripheral blood. Nat. Genet. 46, 430–437 (2014). GWASs and follow-up and an additional n = 1,718/3,055 for the PGS/GRS 21. Cain, A. et al. Multi-cellular communities are perturbed in the aging human brain and with studies). Despite substantial efforts to enhance non-European contribu- Alzheimer’s disease. Preprint at bioRxiv https://doi.org/10.1101/2020.12.22.424084 (2022). tions to GIGASTROKE, we still had limited power for identifying shared 22. Sakaue, S. & Okada, Y. GREP: genome for REPositioning drugs. Bioinformatics 35, 3821–3823 (2019). causal variants through cross-ancestry fine-mapping. We provided 23. Wishart, D. S. et al. DrugBank 5.0: a major update to the DrugBank database for 2018. independent validation of the vast majority of identified genome-wide Nucleic Acids Res. 46, D1074–D1082 (2018). significant associations and graded loci by level of confidence based on 24. Konuma, T., Ogawa, K. & Okada, Y. Integration of genetically regulated gene expression and pharmacological library provides therapeutic drug candidates. Hum. Mol. Genet. 30, these findings. Despite the notable size of the follow-up study sample, 294–304 (2021). with nearly 90,000 additional patients who had a stroke, this analysis 25. Chamorro, A. TP receptor antagonism: a new concept in atherothrombosis and stroke remains underpowered, especially for low-frequency variants and prevention. Cerebrovasc. Dis. 27, 20–27 (2009). 26. Yan, A. et al. Thromboxane A2 receptor antagonist SQ29548 reduces ischemic ancestry- and subtype-specific associations, as most follow-up studies stroke-induced microglia/macrophages activation and enrichment, and ameliorates were derived from large biobanks with event ascertainment based on brain injury. Sci. Rep. 6, 35885 (2016). electronic health records and no suitable stroke subtype information. 27. Bousser, M. G. et al. Terutroban versus aspirin in patients with cerebral ischaemic events (PERFORM): a randomised, double-blind, parallel-group trial. Lancet 377, 2013–2022 The muted risk prediction in clinical-trial participants with previous (2011). stroke history possibly points to the impact of selection or index event 28. Safdar, H. et al. Regulation of the F11, Klkb1, Cyp4v3 gene cluster in livers of metabolically biases and secondary prevention therapy49. challenged mice. PLoS ONE 8, e74637 (2013). 122 | Nature | Vol 611 | 3 November 2022 29. de Haan, H. G. et al. Targeted sequencing to identify novel genetic risk factors for deep 44. Lyden, P. et al. Final results of the RHAPSODY trial: a multi-center, phase 2 trial using a vein thrombosis: a study of 734 genes. J. Thromb. Haemost. 16, 2432–2441 (2018). continual reassessment method to determine the safety and tolerability of 3K3A-APC, 30. Rohmann, J. L. et al. Genetic determinants of activity and antigen levels of contact a recombinant variant of human activated protein C, in Combination with Tissue system factors. J. Thromb. Haemost. 17, 157–168 (2019). plasminogen activator, mechanical thrombectomy or both in moderate to severe acute 31. Backman, J. D. et al. Exome sequencing and analysis of 454,787 UK Biobank participants. ischemic stroke. Ann. Neurol. 85, 125–136 (2019). Nature 599, 628–634 (2021). 45. Huuskonen, M. T. et al. Protection of ischemic white matter and oligodendrocytes in mice 32. Lindstrom, S. et al. Genomic and transcriptomic association studies identify 16 novel by 3K3A-activated protein C. J. Exp. Med. 219, e20211372 (2022). susceptibility loci for venous thromboembolism. Blood 134, 1645–1657 (2019). 46. Chu, W. et al. Blockade of platelet glycoprotein receptor Ib ameliorates blood-brain 33. Verhamme, P. et al. Abelacimab for prevention of venous thromboembolism. N. Engl. J. barrier disruption following ischemic stroke via Epac pathway. Biomed. Pharmacother. Med. 385, 609–617 (2021). 140, 111698 (2021). 34. Weeks E. M. et al. Leveraging polygenic enrichments of gene features to predict genes 47. Yamashita, S. et al. Probucol trial for secondary prevention of atherosclerotic events in underlying complex traits and diseases. Preprint at medRxiv https://doi.org/10.1101/ patients with coronary heart disease (PROSPECTIVE). J. Atheroscler. Thromb. 28, 103–123 2020.09.08.20190561 (2020). (2021). 35. Marston, N. A. et al. Clinical application of a novel genetic risk score for ischemic stroke in 48. Ben-Eghan, C. et al. Don’t ignore genetic data from minority populations. Nature 585, patients with cardiometabolic disease. Circulation 143, 470–478 (2021). 184–186 (2020). 36. Malik, R. et al. Whole-exome sequencing reveals a role of HTRA1 and EGFL8 in brain white 49. Dudbridge, F. et al. Adjustment for index event bias in genome-wide association studies matter hyperintensities. Brain 144, 2670–2682 (2021). of subsequent events. Nat. Commun. 10, 1561 (2019). 37. Söderholm, M. et al. Genome-wide association meta-analysis of functional outcome after ischemic stroke. Neurology 92, e1271–e1283 (2019). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 38. Ma, D. et al. Inhibition of KLF5-Myo9b-RhoA pathway-mediated podosome formation published maps and institutional affiliations. in macrophages ameliorates abdominal aortic aneurysm. Circ. Res. 120, 799–815 (2017). Open Access This article is licensed under a Creative Commons Attribution 39. Yang, X. et al. FURIN expression in vascular endothelial cells is modulated by a coronary 4.0 International License, which permits use, sharing, adaptation, distribution artery disease-associated genetic variant and influences monocyte transendothelial and reproduction in any medium or format, as long as you give appropriate migration. J. Am. Heart Assoc. 9, e014333 (2020). credit to the original author(s) and the source, provide a link to the Creative Commons license, 40. Yakala, G. K. et al. FURIN inhibition reduces vascular remodeling and atherosclerotic and indicate if changes were made. The images or other third party material in this article are lesion progression in mice. Arterioscler. Thromb. Vasc. Biol. 39, 387–401 (2019). included in the article’s Creative Commons license, unless indicated otherwise in a credit line 41. Cantuti-Castelvetri, L. et al. Neuropilin-1 facilitates SARS-CoV-2 cell entry and infectivity. to the material. If material is not included in the article’s Creative Commons license and your Science 370, 856–860 (2020). intended use is not permitted by statutory regulation or exceeds the permitted use, you will 42. Nannoni, S., de Groot, R., Bell, S. & Markus, H. S. Stroke in COVID-19: a systematic review need to obtain permission directly from the copyright holder. To view a copy of this license, and meta-analysis. Int. J. Stroke 16, 137–149 (2021). visit http://creativecommons.org/licenses/by/4.0/. 43. Lyden, P. et al. Phase 1 safety, tolerability and pharmacokinetics of 3K3A-APC in healthy adult volunteers. Curr. Pharm. Des. 19, 7479–7485 (2013). © The Author(s) 2022, corrected publication 2022 Nature | Vol 611 | 3 November 2022 | 123 Article Aniket Mishra1,418, Rainer Malik2,418, Tsuyoshi Hachiya3,418, Tuuli Jürgenson4,5,418, School of Frontier Sciences, The University of Tokyo, Tokyo, Japan. 12Center for Genomic Medicine, Massachusetts General Hospital, Boston, MA, USA. 13Program in Medical and Shinichi Namba , Daniel C. Posner 6,418 , Frederick K. Kamanu , Masaru Koido 7,418 8,9 , 10,11 Population Genetics, Broad Institute of Harvard and the Massachusetts Institute of Quentin Le Grand1, Mingyang Shi11, Yunye He11, Marios K. Georgakis2,12,13, Technology, Cambridge, MA, USA. 14Laboratory of Clinical Genome Sequencing, Department Ilana Caro1, Kristi Krebs4, Yi-Ching Liaw14,15, Felix C. Vaura16,17, Kuang Lin18, of Computational Biology and Medical Sciences, Graduate School of Frontier Sciences, The Bendik Slagsvold Winsvold19,20,21, Vinodh Srinivasasainagendra22, Livia Parodi12,13, University of Tokyo, Tokyo, Japan. 15Department of Public Health and Institute of Public Health, Hee-Joon Bae23, Ganesh Chauhan24, Michael R. Chong25,26, Liisa Tomppo27, Chung Shan Medical University, Taichung, Taiwan. 16Department of Internal Medicine, Rufus Akinyemi28,29, Gennady V. Roshchupkin30,31, Naomi Habib32, Yon Ho Jee33, University of Turku, Turku, Finland. 17Department of Public Health and Welfare, Finnish Institute for Health and Welfare, Turku, Finland. 18Nuffield Department of Population Health, Jesper Qvist Thomassen34, Vida Abedi35,36, Jara Cárcel-Márquez37,38, Marianne Nygaard39,40, University of Oxford, Oxford, UK. 19Department of Research and Innovation, Division of Hampton L. Leonard41,42,43, Chaojie Yang44,45, Ekaterina Yonova-Doing46,47, Maria J. Knol30, Clinical Neuroscience, Oslo University Hospital, Oslo, Norway. 20K. G. Jebsen Center for Adam J. Lewis48, Renae L. Judy49, Tetsuro Ago50, Philippe Amouyel51,52,53, Genetic Epidemiology, Department of Public Health and Nursing, Faculty of Medicine and Nicole D. Armstrong54, Mark K. Bakker55, Traci M. Bartz56,57, David A. Bennett58, Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Joshua C. Bis56, Constance Bordes1, Sigrid Børte20,59,60, Anael Cain32, Paul M. Ridker61,62, Norway. 21Department of Neurology, Oslo University Hospital, Oslo, Norway. 22Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, Birmingham, AL, Kelly Cho7, Zhengming Chen18,63, Carlos Cruchaga64,65, John W. Cole66,67, Phil L. de Jager13,68, USA. 23Department of Neurology and Cerebrovascular Disease Center, Seoul National Rafael de Cid69, Matthias Endres70,71,72,73, Leslie E. Ferreira74, Mirjam I. Geerlings75, University Bundang Hospital, Seoul National University College of Medicine, Seongnam, Natalie C. Gasca57, Vilmundur Gudnason76,77, Jun Hata78, Jing He48, Alicia K. Heath79, Republic of Korea. 24Rajendra Institute of Medical Sciences, Ranchi, India. 25Thrombosis and Yuk-Lam Ho7, Aki S. Havulinna80,81, Jemma C. Hopewell82, Hyacinth I. Hyacinth83, Atherosclerosis Research Institute, David Braley Cardiac, Vascular and Stroke Research Michael Inouye46,84,85,86,87, Mina A. Jacob88, Christina E. Jeon89, Christina Jern90,91, Institute, Hamilton, Ontario, Canada. 26Department of Pathology and Molecular Medicine, Michael G. DeGroote School of Medicine, McMaster University, Hamilton, Ontario, Canada. Masahiro Kamouchi92, Keith L. Keene93, Takanari Kitazono50, Steven J. Kittner67,94, 27 Department of Neurology, Helsinki University Hospital and University of Helsinki, Helsinki, Takahiro Konuma6, Amit Kumar24, Paul Lacaze95, Lenore J. Launer96, Keon-Joo Lee97, Finland. 28Center for Genomic and Precision Medicine, College of Medicine, University of Kaido Lepik4,98,99,100, Jiang Li35, Liming Li101, Ani Manichaikul44, Hugh S. Markus102, Ibadan, Ibadan, Nigeria. 29Neuroscience and Ageing Research Unit Institute for Advanced Nicholas A. Marston8,9, Thomas Meitinger103,104, Braxton D. Mitchell105,106, Medical Research and Training, College of Medicine, University of Ibadan, Ibadan, Nigeria. Felipe A. Montellano107,108, Takayuki Morisaki10, Thomas H. Mosley109, Mike A. Nalls41,42,43, 30 Department of Epidemiology, Erasmus MC University Medical Center Rotterdam, Rotterdam, The Netherlands. 31Department of Radiology and Nuclear Medicine, Erasmus MC Børge G. Nordestgaard , Martin J. O’Donnell , Yukinori Okada 110,111 112 , 6,113,114,115,116,117 University Medical Center Rotterdam, Rotterdam, The Netherlands. 32The Edmond and Lily N. Charlotte Onland-Moret75, Bruce Ovbiagele118, Annette Peters119,120,121, Safra Center for Brain Sciences, The Hebrew University of Jerusalem, Jerusalem, Israel. Bruce M. Psaty56,122,123, Stephen S. Rich44, Jonathan Rosand12,13,124, Marc S. Sabatine8,9, 33 Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, MA, USA. Ralph L. Sacco125,126, Danish Saleheen127, Else Charlotte Sandset128,129, Veikko Salomaa80, 34 Department of Clinical Biochemistry, Copenhagen University Hospital—Rigshospitalet, Muralidharan Sargurupremraj130, Makoto Sasaki3, Claudia L. Satizabal130,131, Copenhagen, Denmark. 35Department of Molecular and Functional Genomics, Weis Center for Research, Geisinger Health System, Danville, VA, USA. 36Department of Public Health Carsten O. Schmidt132, Atsushi Shimizu3, Nicholas L. Smith122,133,134, Kelly L. Sloane135, Sciences, College of Medicine, The Pennsylvania State University, State College, PA, USA. Yoichi Sutoh3, Yan V. Sun136,137, Kozo Tanno3, Steffen Tiedt2, Turgut Tatlisumak138, 37 Stroke Pharmacogenomics and Genetics Laboratory, Biomedical Research Institute Sant Nuria P. Torres-Aguila37, Hemant K. Tiwari22, David-Alexandre Trégouët1, Pau (IIB Sant Pau), Barcelona, Spain. 38Departament de Medicina, Universitat Autònoma de Stella Trompet139,140, Anil Man Tuladhar88, Anne Tybjærg-Hansen34,111, Marion van Vugt141, Barcelona, Barcelona, Spain. 39The Danish Twin Registry, Department of Public Health, Riina Vibo142, Shefali S. Verma143, Kerri L. Wiggins56, Patrik Wennberg144, Daniel Woo83, University of Southern Denmark, Odense, Denmark. 40Department of Clinical Genetics, Odense University Hospital, Odense, Denmark. 41Center for Alzheimer’s and Related Peter W. F. Wilson136,145, Huichun Xu105, Qiong Yang131,146, Kyungheon Yoon147, The COMPASS Dementias, National Institutes of Health, Bethesda, MD, USA. 42Laboratory of Neurogenetics, Consortium*, The INVENT Consortium*, The Dutch Parelsnoer Initiative (PSI) National Institute on Aging, National Institutes of Health, Bethesda, MD, USA. 43Data Tecnica Cerebrovascular Disease Study Group*, The Estonian Biobank*, The PRECISE4Q International, Glen Echo, MD, USA. 44Center for Public Health Genomics, University of Virginia, Consortium*, The FinnGen Consortium*, The NINDS Stroke Genetics Network (SiGN)*, Charlottesville, VA, USA. 45Department of Biochemistry and Molecular Genetics, University of The MEGASTROKE Consortium*, The SIREN Consortium*, The China Kadoorie Biobank Virginia, Charlottesville, VA, USA. 46British Heart Foundation Cardiovascular Epidemiology Unit, Department of Public Health and Primary Care, University of Cambridge, Cambridge, Collaborative Group*, The VA Million Veteran Program*, The International Stroke Genetics UK. 47Department of Genetics, Novo Nordisk Research Centre Oxford, Oxford, UK. Consortium (ISGC)*, The Biobank Japan*, The CHARGE Consortium*, The GIGASTROKE 48 Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN, Consortium*, Iona Y. Millwood18,63, Christian Gieger148, Toshiharu Ninomiya78, USA. 49Department of Surgery, University of Pennsylvania, Philadelphia, PA, USA. Hans J. Grabe149,150, J. Wouter Jukema140,151,152, Ina L. Rissanen75, Daniel Strbian27, 50 Department of Medicine and Clinical Science, Graduate School of Medical Sciences, Young Jin Kim147, Pei-Hsin Chen15, Ernst Mayerhofer12,13, Joanna M. M. Howson46,47, Kyushu University, Fukuoka, Japan. 51University of Lille, INSERM U1167, RID-AGE, LabEx Marguerite R. Irvin54, Hieab Adams153,154, Sylvia Wassertheil-Smoller155, DISTALZ, Risk Factors and Molecular Determinants of Aging-Related Diseases, Lille, France. 52 CHU Lille, Public Health Department, Lille, France. 53Institut Pasteur de Lille, Lille, France. Kaare Christensen39,40,156, Mohammad A. Ikram30, Tatjana Rundek125,126, 54 Department of Epidemiology, University of Alabama at Birmingham, Birmingham, AL, USA. Bradford B. Worrall157,158, G. Mark Lathrop159, Moeen Riaz95, Eleanor M. Simonsick160, 55 UMC Utrecht Brain Center, Department of Neurology and Neurosurgery, University Medical Janika Kõrv142, Paulo H. C. França74, Ramin Zand161,162, Kameshwar Prasad24, Ruth Center Utrecht, University Utrecht, Utrecht, The Netherlands. 56Cardiovascular Health Frikke-Schmidt34,111, Frank-Erik de Leeuw88, Thomas Liman71,75,163, Karl Georg Haeusler108, Research Unit, Department of Medicine, University of Washington, Seattle, WA, USA. Ynte M. Ruigrok55, Peter Ulrich Heuschmann107,164,165, W. T. Longstreth122,166, Keum Ji Jung18,167, 57 Department of Biostatistics, University of Washington, Seattle, WA, USA. 58Rush Alzheimer’s Disease Center, Rush University Medical Center, Chicago, IL, USA. 59Institute of Clinical Lisa Bastarache48, Guillaume Paré25,26,168,169, Scott M. Damrauer170,171, Daniel I. Chasman61,62, Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway. 60Research and Jerome I. Rotter172, Christopher D. Anderson12,13,124,173, John-Anker Zwart19,20,59, Communication Unit for Musculoskeletal Health (FORMI), Department of Research and Teemu J. Niiranen16,17,174, Myriam Fornage175,176, Yung-Po Liaw15,177, Sudha Seshadri130,131,178, Innovation, Division of Clinical Neuroscience, Oslo University Hospital, Oslo, Norway. Israel Fernández-Cadenas37, Robin G. Walters18,63, Christian T. Ruff8,9,419, 61 Division of Preventive Medicine, Brigham and Women’s Hospital, Boston, MA, USA. 62Harvard Mayowa O. Owolabi28,179,419, Jennifer E. Huffman7,419, Lili Milani4,419, Yoichiro Kamatani11,419, Medical School, Boston, MA, USA. 63MRC Population Health Research Unit, University of ✉ ✉ Oxford, Oxford, UK. 64Department of Psychiatry, Washington University School of Medicine, Martin Dichgans2,180,181,419 & Stephanie Debette1,182,419 Saint Louis, MO, USA. 65NeuroGenomics and Informatics Center, Washington University School of Medicine, Saint Louis, MO, USA. 66VA Maryland Health Care System, Baltimore, MD, Bordeaux Population Health Research Center, University of Bordeaux, Inserm, UMR 1219, 1 USA. 67Department of Neurology, University of Maryland School of Medicine, Baltimore, MD, Bordeaux, France. 2Institute for Stroke and Dementia Research (ISD), University Hospital, LMU USA. 68Center for Translational and Computational Neuroimmunology, Department of Munich, Munich, Germany. 3Iwate Tohoku Medical Megabank Organization, Iwate Medical Neurology, Columbia University Medical Center, New York, NY, USA. 69GenomesForLife— University, Iwate, Japan. 4Estonian Genome Centre, Institute of Genomics, University of Tartu, GCAT Lab Group, Germans Trias i Pujol Research Institute (IGTP), Badalona, Spain. 70Klinik und Tartu, Estonia. 5Institute of Mathematics and Statistics, University of Tartu, Tartu, Estonia. Hochschulambulanz für Neurologie, Charité—Universitätsmedizin Berlin, Berlin, Germany. 6 Department of Statistical Genetics, Osaka University Graduate School of Medicine, Suita, 71 Center for Stroke Research Berlin, Berlin, Germany. 72German Center for Neurodegenerative Japan. 7Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC), Diseases (DZNE), partner site Berlin, Berlin, Germany. 73 German Centre for Cardiovascular VA Boston Healthcare System, Boston, MA, USA. 8TIMI Study Group, Boston, MA, USA. Research (DZHK), partner site Berlin, Berlin, Germany. 74Post-Graduation Program on Health 9 Division of Cardiovascular Medicine, Brigham and Women’s Hospital, Harvard Medical and Environment, Department of Medicine and Joinville Stroke Biobank, University of the School, Boston, MA, USA. 10Division of Molecular Pathology, Institute of Medical Sciences, Region of Joinville, Santa Catarina, Brazil. 75Department of Epidemiology, Julius Center for The University of Tokyo, Tokyo, Japan. 11Laboratory of Complex Trait Genomics, Graduate Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht University, Utrecht, The Netherlands. 76Icelandic Heart Association, Kopavogur, Iceland. 77Faculty of Diseases, University of Texas Health Sciences Center, San Antonio, TX, USA. 131Framingham Medicine, University of Iceland, Reykjavik, Iceland. 78Department of Epidemiology and Public Heart Study, Framingham, MA, USA. 132University Medicine Greifswald, Institute for Health, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan. Community Medicine, SHIP/KEF, Greifswald, Germany. 133Kaiser Permanente Washington 79 Department of Epidemiology and Biostatistics, School of Public Health, Imperial College Health Research Institute, Kaiser Permanente Washington, Seattle, WA, USA. 134Department of London, London, UK. 80Department of Public Health and Welfare, Finnish Institute for Health Veterans Affairs Office of Research and Development, Seattle Epidemiologic Research and and Welfare, Helsinki, Finland. 81Institute for Molecular Medicine Finland, FIMM-HiLIFE, Information Center, Seattle, WA, USA. 135Department of Neurology, University of Pennsylvania, Helsinki, Finland. 82Clinical Trial Service and Epidemiological Studies Unit (CTSU), Nuffield Philadelphia, PA, USA. 136Atlanta VA Health Care System, Decatur, GA, USA. 137Department of Department of Population Health, University of Oxford, Oxford, UK. 83Department of Epidemiology, Emory University Rollins School of Public Health, Atlanta, GA, USA. Neurology and Rehabilitation Medicine, University of Cincinnati College of Medicine, 138 Department of Clinical Neuroscience, Institute of Neuroscience and Physiology, Cincinnati, OH, USA. 84Cambridge Baker Systems Genomics Initiative, Department of Public Sahlgrenska Unviersity Hospital, Gothenburg, Sweden. 139Department of Internal Medicine, Health and Primary Care, University of Cambridge, Cambridge, UK. 85Cambridge Baker Section of Gerontology and Geriatrics, Leiden University Medical Center, Leiden, The Systems Genomics Initiative, Baker Heart and Diabetes Institute, Melbourne, Victoria, Netherlands. 140Department of Cardiology, Leiden University Medical Center, Leiden, The Australia. 86Health Data Research UK Cambridge, Wellcome Genome Campus and University Netherlands. 141Division Heart & Lungs, Department of Cardiology, University Medical Center of Cambridge, Cambridge, UK. 87British Heart Foundation Centre of Research Excellence, Utrecht, Utrecht University, Utrecht, The Netherlands. 142Department of Neurology and University of Cambridge, Cambridge, UK. 88Department of Neurology, Donders Center for Neurosurgery, University of Tartu, Tartu, Estonia. 143Department of Pathology and Laboratory Medical Neuroscience, Radboud University Medical Center, Nijmegen, The Netherlands. Medicine, University of Pennsylvania, Philadelphia, PA, USA. 144Department of Public Health 89 Los Angeles County Department of Public Health, Los Angeles, CA, USA. 90Institute of and Clinical Medicine, Umeå University, Umeå, Sweden. 145Department of Medicine, Division Biomedicine, Department of Laboratory Medicine, the Sahlgrenska Academy, University of of Cardiovascular Disease, Emory University School of Medicine, Atlanta, GA, USA. Gothenburg, Gothenburg, Sweden. 91Department of Clinical Genetics and Genomics, 146 Department of Biostatistics, Boston University School of Public Health, Boston, MA, USA. Sahlgrenska University Hospital, Gothenburg, Sweden. 92Department of Health Care 147 Division of Genome Science, Department of Precision Medicine, National Institute of Administration and Management, Graduate School of Medical Sciences, Kyushu University, Health, Cheongju, Republic of Korea. 148Research Unit Molecular Epidemiology, Institute of Fukuoka, Japan. 93Department of Biology, Brody School of Medicine Center for Health Epidemiology, Helmholtz Zentrum München, German Research Center for Environmental Disparities, East Carolina University, Greenville, NC, USA. 94Department of Neurology and Health, Neuherberg, Germany. 149Department of Psychiatry and Psychotherapy, University Geriatric Research and Education Clinical Center, VA Maryland Health Care System, Medicine Greifswald, Greifswald, Germany. 150German Center for Neurodegenerative Baltimore, MD, USA. 95Department of Epidemiology and Preventive Medicine, School of Diseases (DZNE), site Rostock/Greifswald, Rostock, Germany. 151Netherlands Heart Institute, Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, Australia. Utrecht, The Netherlands. 152Einthoven Laboratory for Experimental Vascular Medicine, 96 Intramural Research Program, National Institute on Aging, NIH, Baltimore, MD, USA. LUMC, Leiden, The Netherlands. 153Department of Clinical Genetics, Department of Radiology 97 Department of Neurology, Korea University Guro Hospital, Seoul, Republic of Korea. and Nuclear Medicine, Erasmus MC, Rotterdam, The Netherlands. 154Latin American Brain 98 Department of Computational Biology, University of Lausanne, Lausanne, Switzerland. Health (BrainLat), Universidad Adolfo Ibáñez, Santiago, Chile. 155Department of Epidemiology 99 Swiss Institute of Bioinformatics, Lausanne, Switzerland. 100University Center for Primary and Population Health, Albert Einstein College of Medicine, New York, NY, USA. 156Department Care and Public Health, Lausanne, Switzerland. 101Department of Epidemiology and of Clinical Biochemistry and Pharmacology, Odense University Hospital, Odense, Denmark. Biostatistics, School of Public Health, Peking University Health Science Center, Beijing, China. 157 Department of Neurology, University of Virginia, Charlottesville, VA, USA. 158Department of 102 Stroke Research Group, Department of Clinical Neurosciences, University of Cambridge, Public Health Science, University of Virginia, Charlottesville, VA, USA. 159McGill Genome Cambridge, UK. 103Institute of Human Genetics, Technical University of Munich, Munich, Centre, Montreal, Quebec, Canada. 160Longitudinal Studies Section, Translational Germany. 104Institute of Human Genetics, Helmholtz Zentrum München, German Research Gerontology Branch, National Institute on Aging, Baltimore, MD, USA. 161Geisinger Center for Environmental Health, Neuherberg, Germany. 105Department of Medicine, Neuroscience Institute, Geisinger Health System, Danville, PA, USA. 162Department of University of Maryland School of Medicine, Baltimore, MD, USA. 106Geriatrics Research and Neurology, College of Medicine, The Pennsylvania State University, State College, PA, USA. Education Clinical Center, Baltimore Veterans Administration Medical Center, Baltimore, MD, 163 Klinik für Neurologie, Carl von Ossietzky University of Oldenburg, Oldenburg, Germany. USA. 107Institute of Clinical Epidemiology and Biometry, University of Würzburg, Würzburg, 164 Comprehensive Heart Failure Center, University Hospital Würzburg, Würzburg, Germany. Germany. 108Department of Neurology, University Hospital Würzburg, Würzburg, Germany. 165 Clinical Trial Center, University Hospital Würzburg, Würzburg, Germany. 166Department of 109 The MIND Center, University of Mississippi Medical Center, Jackson, MS, USA. 110Department Neurology, University of Washington, Seattle, WA, USA. 167Institute for Health Promotion, of Clinical Biochemistry, Copenhagen University Hospital—Herlev and Gentofte, Graduate School of Public Health, Yonsei University, Seoul, Republic of Korea. 168Department Copenhagen, Denmark. 111Department of Clinical Medicine, University of Copenhagen, of Health Research Methods, Evidence and Impact, McMaster University, Hamilton, Ontario, Copenhagen, Denmark. 112College of Medicine Nursing and Health Science, NUI Galway, Canada. 169 Population Health Research Institute, David Braley Cardiac, Vascular and Stroke Galway, Ireland. 113Department of Genome Informatics, Graduate School of Medicine, The Research Institute, Hamilton, Ontario, Canada. 170Department of Surgery and Department of University of Tokyo, Tokyo, Japan. 114Laboratory for Systems Genetics, RIKEN Center for Genetics, University of Pennsylvania, Philadelphia, PA, USA. 171Corporal Michael Crescenz VA Integrative Medical Sciences, Yokohama, Japan. 115Laboratory of Statistical Immunology, Medical Center, Philadelphia, PA, USA. 172The Institute for Translational Genomics and Immunology Frontier Research Center (WPI-IFReC), Osaka University, Suita, Japan. Population Sciences, Department of Pediatrics, The Lundquist Institute for Biomedical 116 Integrated Frontier Research for Medical Science Division, Institute for Open and Innovation at Harbor-UCLA Medical Center, Torrance, CA, USA. 173Department of Neurology, Transdisciplinary Research Initiatives, Osaka University, Suita, Japan. 117Center for Infectious Brigham and Women’s Hospital, Boston, MA, USA. 174Division of Medicine, Turku University Disease Education and Research (CiDER), Osaka University, Suita, Japan. 118Weill Institute for Hospital, Turku, Finland. 175Brown Foundation Institute of Molecular Medicine, McGovern Neurosciences, University of California, San Francisco, San Francisco, CA, USA. 119Institute of Medical School, University of Texas Health Science Center at Houston, Houston, TX, USA. Epidemiology, Helmholtz Zentrum München,, German Research Center for Environmental 176 Human Genetics Center, School of Public Health, University of Texas Health Science Health, Neuherberg, Germany. 120Institute for Medical Information Processing, Biometry and Center at Houston, Houston, TX, USA. 177Department of Medical Imaging, Chung Shan Epidemiology, Ludwig Maximilian University Munich, Munich, Germany. 121German Centre for Medical University Hospital, Taichung, Taiwan. 178Department of Neurology, Boston University Cardiovascular Research (DZHK), partner site Munich, Munich, Germany. 122Department of School of Medicine, Boston, MA, USA. 179Department of Medicine, University of Ibadan, Epidemiology, University of Washington, Seattle, WA, USA. 123Department of Health Systems Ibadan, Nigeria. 180Munich Cluster for Systems Neurology, Munich, Germany. 181German and Population Health, University of Washington, Seattle, WA, USA. 124McCance Center for Center for Neurodegenerative Diseases (DZNE), Munich, Germany. 182Department of Brain Health, Massachusetts General Hospital, Boston, MA, USA. 125Department of Neurology, Neurology, Institute for Neurodegenerative Diseases, CHU de Bordeaux, Bordeaux, University of Miami Miller School of Medicine, Miami, FL, USA. 126Evelyn F. McKnight Brain France. 418These authors contributed equally: Aniket Mishra, Rainer Malik, Tsuyoshi Hachiya, Institute, Gainesville, FL, USA. 127Division of Cardiology, Department of Medicine, Columbia Tuuli Jürgenson, Shinichi Namba, Daniel C. Posner. 419These authors jointly supervised University, New York, NY, USA. 128Stroke Unit, Department of Neurology, Oslo University this work: Christian T. Ruff, Mayowa O. Owolabi, Jennifer E. Huffman, Lili Milani, Yoichiro Hospital, Oslo, Norway. 129Research and Development, The Norwegian Air Ambulance Kamatani, Martin Dichgans, Stephanie Debette. *A list of authors and their affiliations Foundation, Oslo, Norway. 130Glenn Biggs Institute for Alzheimer’s and Neurodegenerative appear online. Article The COMPASS Consortium Vascular Emergencies, Department of Emergency Medicine, Massachusetts General Hospital, Boston, MA, USA. 230Department of Emergency Medicine, Harvard Medical School, Boston, Keith L. Keene93, Hyacinth I. Hyacinth83, Joshua C. Bis56, Steven J. Kittner67,94, MA, USA. Braxton D. Mitchell105,106, Guillaume Paré25,26,168,169, Michael R. Chong25,26, Stephen S. Rich44, Mike A. Nalls41,42,43, Hugh S. Markus102, Thomas H. Mosley109, Daniel Woo83, The Dutch Parelsnoer Initiative (PSI) Cerebrovascular Disease Study Group W. T. Longstreth122,166, Bruce M. Psaty56,122,123, Ralph L. Sacco125,126, Tatjana Rundek125,126, Jin-Moo Lee183, Carlos Cruchaga64,65, Martin Dichgans2,180,181,419, Rainer Malik2,418, Mark K. Bakker55, Ynte M. Ruigrok55, Ewoud J. van Dijk88, Peter J. Koudstaal231, Bradford B. Worrall157,158, Myriam Fornage175,176, Yu-Ching Cheng184, Martin J. O’Donnell112, Gert-Jan Luijckx232, Paul J. Nederkoorn233, Robert J. van Oostenbrugge234, James F. Meschia185, Wei Min Chen44, Michèle M. Sale44, Alan B. Zonderman186, Marieke C. Visser233, Marieke J. H. Wermer235 & L. Jaap Kappelle55 Michele K. Evans186, James G. Wilson187, Adolfo Correa187, Matthew Traylor188, Cathryn M. Lewis189, Cara L. Carty190, Alexander Reiner122,191, Jeffrey Haessler191, Carl D. Langefeld192, Rebecca F. Gottesman193,194,195, Kristine Yaffe196, Yong Mei Liu192, Department of Neurology, Erasmus University Medical Center, Rotterdam, The Netherlands. 231 Charles Kooperberg191, Leslie A. Lange197, Karen L. Furie198, Donna K. Arnett199, Department of Neurology, University Medical Center Groningen, Groningen, The 232 Oscar R. Benavente200, Raji P. Grewal201 & Leema Reddy Peddareddygari201 Netherlands. 233Department of Neurology, Amsterdam UMC, University of Amsterdam, Amsterdam, The Netherlands. 234Department of Neurology, Cardiovascular Research Institute Maastricht, Maastricht University Medical Center, Maastricht, The Netherlands. 235Department 183 Department of Neurology, Washington University School of Medicine, Saint Louis, MO, USA. of Neurology, Leiden University Medical Center, Leiden, The Netherlands. 184 Baltimore Veterans Administration Medical Center and University of Maryland School of Medicine, Baltimore, MD, USA. 185Mayo Clinic Florida, Jacksonville, FL, USA. 186Laboratory of Epidemiology and Population Science, National Institute on Aging, National Institutes of The Estonian Biobank Health, Baltimore, MD, USA. 187University of Mississippi Medical Center, Jackson, MS, USA. 188 William Harvey Research Institute, Barts and The London School of Medicine and Dentistry, Tuuli Jürgenson4,5,418, Kristi Krebs4, Kaido Lepik4,98,99,100, Lili Milani4,419, Tõnu Esko4, Queen Mary University of London, London, UK. 189Social, Genetic and Developmental Andres Metspalu4, Reedik Mägi4 & Mari Nelis4 Psychiatry Centre, King’s College London, London, UK. 190Initiative for Research and Education to Advance Community Health, Washington State University, Seattle, WA, USA. 191 Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA, USA. 192Division of Public Health Sciences, Wake Forest School of Medicine, Winston-Salem, The PRECISE4Q Consortium NC, USA. 193Johns Hopkins University School of Medicine, Baltimore, MD, USA. 194Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD, USA. 195Stroke Lili Milani4,419 Branch, National Institute of Neurological Disorders and Stroke, Bethesda, MD, USA. 196 University of California, San Francisco, San Francisco, CA, USA. 197University of Colorado Anschutz Medical Campus, Denver, CO, USA. 198Brown University Warren Alpert Medical School, Providence, RI, USA. 199College of Public Health, University of Kentucky, Lexington, The FinnGen Consortium KY, USA. 200University of British Columbia, Vancouver, British Columbia, Canada. 201 Neuroscience Institute, Saint Francis Medical Center, Trenton, NJ, USA. Felix C. Vaura16,17Teemu J Niiranen16,17,174Aki S. Havulinna80,81 The INVENT Consortium The NINDS Stroke Genetics Network (SiGN) Daniel I. Chasman61,62, Kerri L. Wiggins56, Jerome I. Rotter172, Bruce M. Psaty56,122,123, Charles Kooperberg191, Kristian Hveem20,202,203, Paul M. Ridker61,62, Rainer Malik2,418, Michael R. Chong25,26, Liisa Tomppo27, Vida Abedi35,36, Carlos Cruchaga64,65, David-Alexandre Trégouët1, Nicholas L. Smith122,133,134, Sara Lindstrom122,191, Lu Wang204, John W. Cole66,67, Christina Jern90,91, Steven J. Kittner67,94, Hugh S. Markus102, Erin N. Smith205,206, William Gordon122, Astrid van Hylckama Vlieg207, Mariza de Andrade208, Braxton D. Mitchell105,106, Jonathan Rosand12,13,124, Ralph L. Sacco125,126, Danish Saleheen127, Jennifer A. Brody56, Jack W. Pattee209, Jeffrey Haessler191, Ben M. Brumpton20,210,211, Turgut Tatlisumak138, Daniel Woo83, Huichun Xu105, Christian Gieger148, Daniel Strbian27, Pierre Suchon212,213, Ming-Huei Chen131, Kelly A. Frazer205,206,214, Constance Turman215, Marguerite R. Irvin54, Sylvia Wassertheil-Smoller155, Tatjana Rundek125,126, Marine Germain1, James MacDonald204, Sigrid K. Braekkan206,216, Sebastian M. Armasu208, Bradford B. Worrall157,158, Ramin Zand161,162, Frank-Erik de Leeuw88, Guillaume Paré25,26,168,169, Nathan Pankratz217, Rebecca D. Jackson218,219, Jonas B. Nielsen20,220,221, Franco Giulianini61, Israel Fernández-Cadenas37, Yoichiro Kamatani11,419, Martin Dichgans2,180,181,419, Marja K. Puurunen131, Manal Ibrahim212, Susan R. Heckbert122,133, Theo K. Bammler204, Stephanie Debette1,182,419, Christopher R. Levi236, Jane Maguire237, Jordi Jiménez-Conde238, Bryan M. McCauley208, Kent D. Taylor172,222,223, James S. Pankow224, Alexander P. Reiner122,191, Pankaj Sharma239,240, Cathie L. M. Sudlow241,242, Kristiina Rannikmäe241,242, Maiken E. Gabrielsen20, Jean-François Deleuze225,226, Chris J. O’Donnell131,227, Jihye Kim215, Reinhold Schmidt243, James F. Meschia185, Agnieszka Slowik244, Joanna Pera244, Barbara McKnight57,191, Peter Kraft215, John-Bjarne Hansen206,216, Frits R. Rosendaal207, Vincent N. S. Thijs245,246, Arne G. Lindgren247,248, Andreea Ilinca247,248, Olle Melander249, John A. Heit208, Weihong Tang224, Pierre-Emmanuel Morange212,213,228, Andrew D. Johnson131 & Gunnar Engström249, Raji P. Grewal201, Leema Reddy Peddareddygari201, Christopher Kabrhel229,230 Kathryn M. Rexrode250, Peter M. Rothwell251, Leslie A. Lange197, Tara M. Stanne90, Julie A. Johnson252,253, John Danesh254,255,256,257, Adam S. Butterworth254,258, Jin-Moo Lee183, 202 HUNT Research Center, Department of Public Health and Nursing, Faculty of Medicine and Laura Heitsch183,259, Giorgio B. Boncoraglio260, Michiaki Kubo261, Alessandro Pezzini262, Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Arndt Rolfs263, Anne-Katrin Giese264, David Weir265, Rebecca D. Jackson218,219, Norway. 203Department of Research, Innovation and Education, St Olavs Hospital, Trondheim Owen A. Ross185, Robin Lemmons266,267, Martin Soderholm248,249, Mary Cushman268, University Hospital, Trondheim, Norway. 204Department of Environmental and Occupational Katarina Jood269,270, Caitrin W. McDonough252, Steven Bell102, Birgit Linkohr119, Health Sciences, University of Washington, Seattle, WA, USA. 205Department of Pediatrics and Tsong-Hai Lee271 & Jukka Putaala27 Rady Children’s Hospital, University of California San Diego, La Jolla, CA, USA. 206K. G. Jebsen Thrombosis Research and Expertise Center, Department of Clinical Medicine, UiT—The Arctic 236 John Hunter Hospital, Hunter Medical Research Institute and University of Newcastle, University of Norway, Tromsø, Norway. 207Department of Clinical Epidemiology, Leiden Newcastle, New South Wales, Australia. 237Faculty of Health, University of Technology Sydney, University Medical Center, Leiden, The Netherlands. 208Department of Health Sciences Ultimo, New South Wales, Australia. 238Department of Neurology, IMIM-Hospital del Mar, Research, Mayo Clinic, Rochester, MN, USA. 209Division of Biostatistics, School of Public Neurovascular Research Group, IMIM (Institut Hospital del Mar d’Investigacions Mèdiques), Health, University of Minnesota, Minneapolis, MN, USA. 210MRC Integrative Epidemiology Unit, Universitat Autònoma de Barcelona/DCEXS-Universitat Pompeu Fabra, Barcelona, Spain. University of Bristol, Bristol, UK. 211Clinic of Thoracic and Occupational Medicine, St Olavs 239 Institute of Cardiovascular Research, Royal Holloway University of London, London, UK. Hospital, Trondheim University Hospital, Trondheim, Norway. 212Laboratory of Haematology, 240 Ashford and St Peters Hospital, Chertsey, UK. 241Usher Institute of Population Health La Timone Hospital, Marseille, France. 213C2VN, University of Aix Marseille, INSERM, INRAE, Sciences and Informatics, University of Edinburgh, Edinburgh, UK. 242Centre for Clinical Brain C2VN, Marseille, France. 214Institute of Genomic Medicine, University of California, San Diego, Sciences, University of Edinburgh, Edinburgh, UK. 243Department of Neurology, Medical La Jolla, CA, USA. 215Program in Genetic Epidemiology and Statistical Genetics, Harvard T. H. University of Graz, Graz, Austria. 244Department of Neurology, Jagiellonian University, Krakow, Chan School of Public Health, Boston, MA, USA. 216Division of Internal Medicine, University Poland. 245Stroke Division, Florey Institute of Neuroscience and Mental Health, University of Hospital of North Norway, Tromsø, Norway. 217Department of Laboratory Medicine and Melbourne, Heidelberg, Melbourne, Victoria, Australia. 246Department of Neurology, Austin Pathology, School of Medicine, University of Minnesota, Minneapolis, MN, USA. 218Division of Health, Heidelberg, Victoria, Australia. 247Department of Clinical Sciences Lund, Neurology, Endocrinology, Diabetes and Metabolism, The Ohio State University, Columbus, OH, USA. Lund University, Lund, Sweden. 248Department of Neurology, Rehabilitation Medicine, Memory 219 Department of Internal Medicine and the Center for Clinical and Translational Science, Ohio Disorders, Geriatrics, Skåne University Hospital, Lund, Sweden. 249Department of Clinical State University, Columbus, OH, USA. 220Department of Internal Medicine, Division of Sciences, Lund University, Malmö, Sweden. 250Department of Medicine, Brigham and Women’s Cardiology, University of Michigan, Ann Arbor, MI, USA. 221Department of Epidemiology Hospital, Boston, MA, USA. 251Wolfson Centre for Prevention of Stroke and Dementia, Nuffield Research, Statens Serum Institut, Copenhagen, Denmark. 222Institute for Translational Department of Clinical Neurosciences, University of Oxford, Oxford, UK. 252Department of Genomics and Population Sciences, Los Angeles Biomedical Research Institute at Harbor– Pharmacotherapy and Translational Research and Center for Pharmacogenomics, College of UCLA Medical Center, Torrance, CA, USA. 223Division of Genomic Outcomes, Department of Pharmacy, University of Florida, Gainesville, FL, USA. 253Division of Cardiovascular Medicine, Pediatrics, Harbor–UCLA Medical Center, Torrance, CA, USA. 224Division of Epidemiology and College of Medicine, University of Florida, Gainesville, FL, USA. 254BHF Cardiovascular Community Health, School of Public Health, University of Minnesota, Minneapolis, MN, USA. Epidemiology Unit, Department of Public Health and Primary Care, University of Cambridge, 225 Centre National de Recherche en Génomique Humaine, Direction de la Recherche Cambridge, UK. 255NIHR Blood and Transplant Research Unit in Donor Health and Genomics, Fondamentale, CEA, Evry, France. 226CEPH-Fondation Jean Dausset, Paris, France. 227Million Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK. Veteran’s Program, Veteran’s Administration, Boston, MA, USA. 228CRB Assistance Publique— 256 Wellcome Trust Sanger Institute, Cambridge, UK. 257British Heart Foundation Cambridge Hôpitaux de Marseille, HemoVasc (CRB AP-HM HemoVasc), Marseille, France. 229Center for Centre of Excellence, Department of Medicine, University of Cambridge, Cambridge, UK. 258 National Institute for Health Research Blood and Transplant Research Unit in Donor Health 274 Stroke Pharmacogenomics and Genetics Laboratory, Fundación Docència I Recerca and Genomics, University of Cambridge, Cambridge, UK. 259Department of Emergency Mútua Terrassa, Hospital Mútua Terrassa, Terrassa, Spain. 275Neurovascular Research Medicine, Washington University School of Medicine, Saint Louis, MO, USA. 260Department of Laboratory, Vall d’Hebron Institute of Research, Universitat Autònoma de Barcelona, Cerebrovascular Diseases, Fondazione IRCCS Istituto Neurologico ‘Carlo Besta’, Milan, Italy. Barcelona, Spain. 276Neurology Service, Hospital Universitari Mútua Terrassa, Terrassa, 261 Laboratory for Statistical Analysis, RIKEN Center for Integrative Medical Sciences, Spain. 277Department of Neurology, Beth Israel Deaconess Medical Center, Boston, MA, USA. Yokohama, Japan. 262Department of Clinical and Experimental Sciences, Neurology Clinic, 278 Institute de Biomedicine of Seville, IBiS/Hospital Universitario Virgen del Rocío/CSIC/ University of Brescia, Brescia, Italy. 263Albrecht Kossel Institute, University Clinic of Rostock, University of Seville and Department of Neurology, Hospital Universitario Virgen Macarena, Rostock, Germany. 264Department of Neurology, Massachusetts General Hospital (MGH), Seville, Spain. 279Department of Epidemiology, Human Genetics, and Environmental Harvard Medical School, Boston, MA, USA. 265Survey Research Center, University of Michigan, Sciences (EHGES), UTHealth Science Center, School of Public Health, Houston, TX, USA. Ann Arbor, MI, USA. 266Department of Neurosciences, Experimental Neurology, KU Leuven– 280 BHF Glasgow Cardiovascular Research Centre, Faculty of Medicine, University of University of Leuven, Leuven, Belgium. 267Department of Neurology, VIB Center for Brain & Glasgow, Glasgow, UK. 281German Centre for Cardiovascular Research (DZHK), partner site Disease Research, University Hospitals Leuven, Leuven, Belgium. 268Department of Medicine, Greifswald, Greifswald, Germany. 282Program in Bioinformatics and Integrative Genomics, Larner College of Medicine at the University of Vermont, Burlington, VT, USA. 269Department of Harvard Medical School, Boston, MA, USA. 283Department of Statistical Genetics, Osaka Clinical Neuroscience, Institute of Neuroscience and Physiology, Sahlgrenska Academy at University Graduate School of Medicine, Osaka, Japan. 284deCODE genetics/AMGEN, University of Gothenburg, Gothenburg, Sweden. 270Region Västra Götaland, Department of Reykjavik, Iceland. 285J. Philip Kistler Stroke Research Center, Department of Neurology, Neurology, Sahlgrenska University Hospital, Gothenburg, Sweden. 271Chang Gung Memorial MGH, Boston, MA, USA. 286Population Health Research Institute, McMaster University, Hospital, Linkou Medical Center in Taiwan, Taoyuan City, Taiwan. Hamilton, Ontario, Canada. 287AA Martinos Center for Biomedical Imaging, Department of Radiology, MGH, Harvard Medical School, Boston, MA, USA. 288School of Life Science, The MEGASTROKE Consortium University of Lincoln, Lincoln, UK. 289Department of Neurology, Mayo Clinic, Rochester, MN, USA. 290Stroke Pharmacogenomics and Genetics, Fundacio Docència i Recerca MutuaTerrassa, Terrassa, Spain. 291Department of Medical Genetics, University Medical Aniket Mishra1,418, Rainer Malik2,418, Ganesh Chauhan24, Michael R. Chong25,26, Liisa Tomppo27, Center Utrecht, Utrecht, The Netherlands. 292Boston University School of Public Health, Tetsuro Ago50, Philippe Amouyel51,52,53, Traci M. Bartz56,57, Paul M. Ridker61,62, Boston, MA, USA. 293The Beijer Laboratory and Department of Immunology, Genetics and Carlos Cruchaga64,65, John W. Cole66,67, Vilmundur Gudnason76,77, Jun Hata78, Pathology, Uppsala University and Science for Life Laboratory, Uppsala, Sweden. Aki S. Havulinna80,81, Jemma C. Hopewell82, Hyacinth I. Hyacinth83, Christina Jern90,91, 294 Department of Genetics, University of North Carolina, Chapel Hill, NC, USA. 295Department Masahiro Kamouchi92, Keith L. Keene93, Takanari Kitazono50, Steven J. Kittner67,94, of Neurology and Stroke Center, Basel University Hospital, Basel, Switzerland. Ani Manichaikul44, Hugh S. Markus102, Braxton D. Mitchell105,106, Thomas H. Mosley109, 296 Neurorehabilitation Unit, University of Basel and University Center for Medicine of Aging Mike A. Nalls41,42,43, Martin J. O’Donnell112, Yukinori Okada6,113,114,115,116,117, Bruce M. Psaty56,122,123, and Rehabilitation Basel, Felix Platter Hospital, Basel, Switzerland. 297Department of Stephen S. Rich44, Jonathan Rosand12,13,124, Ralph L. Sacco125,126, Danish Saleheen127, Neurology, Yale University School of Medicine, New Haven, CT, USA. 298Department of Veikko Salomaa80, Muralidharan Sargurupremraj130, Makoto Sasaki3, Medical Sciences, Molecular Epidemiology and Science for Life Laboratory, Uppsala Claudia L. Satizabal130,131, Carsten O. Schmidt132, Kozo Tanno3, Steffen Tiedt2, University, Uppsala, Sweden. 299Department of Neurology, Leeds General Infirmary, Leeds Turgut Tatlisumak138, Nuria P. Torres-Aguila37, Stella Trompet139,140, Kerri L. Wiggins56, Teaching Hospitals NHS Trust, Leeds, UK. 300Public Health Stream, Hunter Medical Research Daniel Woo83, Huichun Xu105, Qiong Yang131,146, Toshiharu Ninomiya78, Institute, New Lambton, New South Wales, Australia. 301College of Health, Medicine and J. Wouter Jukema140,151,152, Daniel Strbian27, Joanna M. M. Howson46,47, Marguerite R. Irvin54, Wellbeing, The University of Newcastle, Newcastle, New South Wales, Australia. Hieab Adams153,154, Sylvia Wassertheil-Smoller155, Mohammad A. Ikram30, 302 Department of Biostatistics and Data Science, Division of Public Health Sciences, Wake Tatjana Rundek125,126, Bradford B. Worrall157,158, W. T. Longstreth122,166, Guillaume Paré25,26,168,169, Forest University School of Medicine, Winston-Salem, NC, USA. 303Department of Medicine, Daniel I. Chasman61,62, Jerome I. Rotter172, Christopher D. Anderson12,13,124,173, Division of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, CA, Myriam Fornage175,176, Sudha Seshadri130,131,178, Israel Fernández-Cadenas37, USA. 304MRC Epidemiology Unit, School of Clinical Medicine, Institute of Metabolic Science, Yoichiro Kamatani11,419, Martin Dichgans2,180,181,419, Stephanie Debette1,182,419, Matthew Traylor188, University of Cambridge, Cambridge, UK. 305 Computational Medicine, Berlin Institute of Oscar L. Lopez272, Cara L. Carty190, Adolfo Correa187, Raji P. Grewal201, Jeffrey Haessler191, Health at Charité – Universitätsmedizin Berlin, Berlin, Germany. 306 Precision Healthcare Susan R. Heckbert122,133, Xueqiu Jian130,131, Jordi Jiménez-Conde238, Charles Kooperberg191, Institute, Queen Mary University of London, London, UK. 307INSERM U 1172, CHU Lille, Leslie A. Lange197, Carl D. Langefeld192, Jin-Moo Lee183, Cathryn M. Lewis189,­ Université Lille, Lille, France. 308MRC Biostatistics Unit, University of Cambridge, Cambridge, James F. Meschia185, Alexander Reiner122,191, Ulf Schminke273, James G. Wilson187, UK. 309Bioinformatics Core Facility, University of Gothenburg, Gothenburg, Sweden. Donna K. Arnett199, Yu-Ching Cheng184, Natalia Cullell37,274, Pilar Delgado275, Laura Ibañez64, 310 Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Jerzy Krupinski276, Vasileios Lioutas131,277, Koichi Matsuda14, Joan Montaner278, Elena Muiño37, Sweden. 311Department of Allergy and Rheumatology, Graduate School of Medicine, Leema Reddy Peddareddygari201, Jaume Roquer238, Chloe Sarnowski279, Naveed Sattar280, University of Tokyo, Tokyo, Japan. 312Department of Pediatrics, College of Medicine, Gerli Sibolt27, Alexander Teumer132,281, Loes Rutten-Jacobs102, Masahiro Kanai261,282,283, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA. 313Department of Anne-Katrin Giese264, Solveig Gretarsdottir284, Natalia S. Rost264,285, Salim Yusuf286, Neurology, Justus Liebig University, Giessen, Germany. 314Department of Public Health, Peter Almgren249, Hakan Ay285,287, Steve Bevan288, Giorgio B. Boncoraglio260, Bordeaux University Hospital, Bordeaux, France. 315School of Medicine, Dentistry and Robert D. Brown Jr289, Adam S. Butterworth254,258, Caty Carrera275,290, Julie E. Buring61,62, Nursing at the University of Glasgow, Glasgow, UK. 316University of Newcastle and Hunter Wei-Min Chen44, Ioana Cotlarciuc239,240, John Danesh254,255,256,257, Paul I. W. de Bakker75,291, Medical Research Institute, New Lambton, New South Wales, Australia. 317INM, University of Anita L. DeStefano131,146,292, Marcel den Hoed293, Qing Duan294, Stefan T. Engelter295,296, Montpellier Inserm U1298, Montpellier, France. 318Centre for Research in Environmental Guido J. Falcone13,297, Rebecca F. Gottesman195, Stefan Gustafsson298, Ahamad Hassan299, Epidemiology, Barcelona, Spain. 319Department of Neurology, Università degli Studi di Elizabeth G. Holliday300,301, George Howard22, Fang-Chi Hsu302, Erik Ingelsson298,303, Perugia, Umbria, Italy. 320Broad Institute, Cambridge, MA, USA. 321Department of Neurology, Tamara B. Harris186, Julie A. Johnson252,253, Brett M. Kissela83, Dawn O. Kleindorfer83, Memory Clinic, Bordeaux University Hospital, Bordeaux, France. 322Department of Internal Michiaki Kubo261, Claudia Langenberg304,305,306, Robin Lemmens266,267, Didier Leys307, Medicine B, University Medicine Greifswald, Greifswald, Germany. 323Robertson Center for Wei-Yu Lin254,308, Arne G. Lindgren247,248, Erik Lorentzen309, Patrik K. Magnusson310, Biostatistics, University of Glasgow, Glasgow, UK. 324Hero DMC Heart Institute, Dayanand Jane Maguire237, Patrick F. McArdle105, Sara L. Pulit55,291, Kristiina Rannikmäe241,242, Medical College & Hospital, Ludhiana, India. 325Atherosclerosis Research Unit, Department Kathryn M. Rexrode250, Kenneth Rice57, Peter M. Rothwell251, Saori Sakaue283,311, of Medicine Solna, Karolinska Institutet, Stockholm, Sweden. 326Tohoku Medical Megabank Bishwa R. Sapkota312, Reinhold Schmidt243, Pankaj Sharma239,240, Agnieszka Slowik244, Organization, Sendai, Japan. 327Department of Public Health and Caring Sciences/Geriatrics, Cathie L. M. Sudlow241,242, Christian Tanislav313, Vincent N. S. Thijs245,246, Uppsala University, Uppsala, Sweden. 328 Krembil Brain Institute, University Health Network, Gudmar Thorleifsson284, Unnur Thorsteinsdottir284, Christophe Tzourio1,314, Cornelia M. van Toronto, Ontario, Canada. 329 Department of Medicine and Tanz Centre for Research in Duijn18,30, Matthew Walters315, Nicholas J. Wareham266, Najaf Amin30, Hugo J. Aparicio131,178, Neurodegenerative Diseases, University of Toronto, Toronto, Ontario, Canada. John Attia316, Alexa S. Beiser131,146, Claudine Berr317, Mariana Bustamante318, Valeria Caso319, 330 Epidemiology and Prevention Group, Center for Public Health Sciences, National Cancer Seung Hoan Choi48,320, Ayesha Chowhan131,178, Jean-François Dartigues1,321, Center, Tokyo, Japan. 331Department of Basic and Clinical Neurosciences, King’s College Hossein Delavaran247,248, Marcus Dörr281,322, Gunnar Engström249, Ian Ford323, London, London, UK. 332Departments of Neurology & Radiology, Landspitali National Wander S. Gurpreet324, Anders Hamsten325, Laura Heitsch183,259, Atsushi Hozawa326, University Hospital, Reykjavik, Iceland. 333Department of Neurology, Heidelberg University Andreea Ilinca247,248, Martin Ingelsson327,328,329, Motoki Iwasaki330, Rebecca D. Jackson218,219, Hospital, Heidelberg, Germany. 334Montefiore Medical Center, Albert Einstein College of Katarina Jood269,270, Sara Kaffashian1, Lalit Kalra331, Olafur Kjartansson332, Manja Kloss333, Medicine, New York, NY, USA. 335Department of Medical Sciences, Uppsala University, Peter J. Koudstaal231, Daniel L. Labovitz334, Cathy C. Laurie57, Christopher R. Levi236, Uppsala, Sweden. 336Genetic and Genomic Epidemiology Unit, Wellcome Trust Centre for Linxin Li251, Lars Lind335, Cecilia M. Lindgren336,337, Yong Mei Liu192, Oscar L. Lopez272, Human Genetics, University of Oxford, Oxford, UK. 337Wellcome Trust Centre for Human Hirata Makoto338, Naoko Minegishi326, Andrew P. Morris337,339, Martina Müller-Nurasy Genetics, Oxford, UK. 338BioBankJapan, Laboratory of Clinical Sequencing, Department of id121,340,341, Bo Norrving247, Soichi Ogishima219, Eugenio A. Parati260, Nancy L. Pedersen310, Computational Biology and Medical Sciences, Graduate School of Frontier Sciences, Joanna Pera244, Markus Perola80, Pekka Jousilahti80, Alessandro Pezzini262, Silvana Pileggi342, University of Tokyo, Tokyo, Japan. 339Department of Biostatistics, University of Liverpool, Raquel Rabionet343, Iolanda Riba-Llena275, Marta Ribasés344, Jose R. Romero131,178, Liverpool, UK. 340Institute of Genetic Epidemiology, Helmholtz Zentrum München—German Anthony G. Rudd345,346, Antti-Pekka Sarin81,130, Ralhan Sarju324, Mamoru Satoh284, Research Center for Environmental Health, Neuherberg, Germany. 341Department of Norie Sawada330, Ásgeir Sigurdsson284, Albert Smith347, Tara M. Stanne90, O. Colin Stine348, Medicine I, Ludwig-Maximilians-Universität, Munich, Germany. 342Translational Genomics David J. Stott349, Konstantin Strauch340,350, Takako Takai326, Hideo Tanaka351,352, Unit, Department of Oncology, IRCCS Istituto di Ricerche Farmacologiche Mario Negri, Emmanuel Touze353,354, Shoichiro Tsugane330, Andre G. Uitterlinden355, Milan, Italy. 343Department of Genetics, Microbiology and Statistics, University of Barcelona, Einar M. Valdimarsson356, Sven J. van der Lee30, Kenji Wakai351, David Weir265, Barcelona, Spain. 344Psychiatric Genetics Unit, Group of Psychiatry, Mental Health and Stephen R. Williams156, Charles D. A. Wolfe345,346, Quenna Wong57, Taiki Yamaji330, Addictions, Vall d’Hebron Research Institute (VHIR), Universitat Autònoma de Dharambir K. Sanghera312,357,358, Olle Melander249, Arndt Rolfs263, Kari Stefansson77,284, Barcelona,Biomedical Network Research Centre on Mental Health (CIBERSAM), Barcelona, Kent D. Taylor172,222,223, Nicolas Martinez-Majander27, Kenji Sobue3, Spain. 345National Institute for Health Research Comprehensive Biomedical Research Carolina Soriano-Tárraga359 & Henry Völzke132 Centre, Guy’s & St Thomas’ NHS Foundation Trust and King’s College London, London, UK. 346 School of Lifecourse and Population Sciences, King’s College London, London, UK. Department of Neurology, School of Medicine, University of Pittsburgh, Pittsburgh, PA, 272 347 Icelandic Heart Association, Reykjavik, Iceland. 348Department of Epidemiology, University USA. 273Department of Neurology, University Medicine Greifswald, Greifswald, Germany. of Maryland School of Medicine, Baltimore, MD, USA. 349Institute of Cardiovascular and Article Medical Sciences, College of Medical, Veterinary and Life Sciences, University of Glasgow, The CHARGE Consortium Glasgow, UK. 350IBE, Faculty of Medicine, LMU Munich, Munich, Germany. 351Division of Epidemiology and Prevention, Aichi Cancer Center Research Institute, Nagoya, Japan. Aniket Mishra1,418, Quentin Le Grand1, Ilana Caro1, Gennady V. Roshchupkin30,31, 352 Department of Epidemiology, Nagoya University Graduate School of Medicine, Nagoya, Hampton L. Leonard41,42,43, Chaojie Yang44,45, Maria J. Knol30, Traci M. Bartz56,57, Japan. 353Department of Neurology, Caen University Hospital, Caen, France. 354University of Joshua C. Bis56, Constance Bordes1, Paul M. Ridker61,62, Mirjam I. Geerlings75, Caen Normandy, Caen, France. 355Department of Internal Medicine, Erasmus University Natalie C. Gasca57, Vilmundur Gudnason76,77, Paul Lacaze95, Lenore J. Launer96, Medical Center, Rotterdam, The Netherlands. 356Landspitali University Hospital, Reykjavik, Ani Manichaikul44, Thomas H. Mosley109, Mike A. Nalls41,42,43, Stephen S. Rich44, Iceland. 357Department of Pharmaceutical Sciences, College of Pharmacy, University of Muralidharan Sargurupremraj130, Claudia L. Satizabal130,131, Carsten O. Schmidt132, Oklahoma Health Sciences Center, Oklahoma City, OK, USA. 358Oklahoma Center for Stella Trompet139,140, Marion van Vugt141, Qiong Yang131,146, Hans J. Grabe149,150, Neuroscience, Oklahoma City, OK, USA. 359IMIM (Hospital del Mar Medical Research J. Wouter Jukema140,151,152, Ina L. Rissanen75, Hieab Adams153,154, Sylvia Wassertheil-Smoller155, Institute), Barcelona, Spain. Mohammad A. Ikram30, Moeen Riaz95, Eleanor M. Simonsick160, W. T. Longstreth122,166, Daniel I. Chasman61,62, Jerome I. Rotter172, Myriam Fornage175,176, Sudha Seshadri130,131,178, Stephanie Debette1,182,419, Rebecca F. Gottesman195, Naveed Sattar280, David J. Stott349, The SIREN Consortium Eric J. Shiroma186, Oscar L. Lopez272, Sigurdur Sigurdsson76, Mohsen Ghanbari30, Ulf Schminke273, Eric Boerwinkle176,377, Hugo J. Aparicio131,178, Alexa S. Beiser131,146, Vinodh Srinivasasainagendra22, Rufus Akinyemi28,29, Bruce Ovbiagele118, Hemant K. Tiwari22, Jose R. Romero131,178, Vasileios Lioutas131,277, Xueqiu Jian130,131, Bernard Fongang130,131, Mayowa O. Owolabi28,179,419, Onoja Akpa360,361, Fred S. Sarfo362, Albert Akpalu363, Ruiqi Wang131,146, Chloe Sarnowski279, Mohammad K. Ikram30,231, Alexander Teumer132,281 & Reginald Obiako364, Kolawole Wahab365, Godwin Osaigbovo366, Lukman Owolabi367, Uwe Völker281,378 Morenikeji Komolafe368, Carolyn Jenkins369, Oyedunni Arulogun370, Godwin Ogbole364, Abiodun M. Adeoye28,179, Joshua Akinyemi360, Atinuke Agunloye371, Adekunle G. Fakunle179, Human Genome Sequencing Center, Baylor College of Medicine, Houston, TX, USA. 377 Ezinne Uvere179, Abimbola Olalere179 & Olayinka J. Adebajo179 Interfaculty Institute for Genetics and Functional Genomics, University Medicine 378 Greifswald, Greifswald, Germany. 360 Department of Epidemiology and Medical Statistics, University of Ibadan, Ibadan, Nigeria. 361 Institute of Cardiovascular Diseases, University of Ibadan, Ibadan, Nigeria. 362Department of Medicine, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana. The GIGASTROKE Consortium 363 Department of Medicine, University of Ghana Medical School, Accra, Ghana. 364Department of Medicine, Ahmadu Bello University, Zaria, Nigeria. 365Department of Medicine, University of Aniket Mishra1,418, Rainer Malik2,418, Tsuyoshi Hachiya3,418, Tuuli Jürgenson4,5,418, Ilorin Teaching Hospital, Ilorin, Nigeria. 366Jos University Teaching Hospital, Jos, Nigeria. Shinichi Namba6,418, Daniel C. Posner7,418, Frederick K. Kamanu8,9, Masaru Koido10,11, 367 Department of Medicine, Aminu Kano Teaching Hospital, Kano, Nigeria. 368Department of Quentin Le Grand1, Mingyang Shi11, Yunye He11, Marios K. Georgakis2,12,13, Medicine, Obafemi Awolowo University Teaching Hospital, Ile-Ife, Nigeria. 369Medical Ilana Caro1, Kristi Krebs4, Yi-Ching Liaw14,15, Felix C. Vaura16,17, Kuang Lin18, University of South Carolina, Charleston, SC, USA. 370Department of Health Promotion and Bendik Slagsvold Winsvold19,20,21, Vinodh Srinivasasainagendra22, Livia Parodi12,13, Education, University of Ibadan, Ibadan, Nigeria. 371Department of Radiology, College of Hee-Joon Bae23, Ganesh Chauhan24, Michael R. Chong25,26, Liisa Tomppo27, Medicine, University of Ibadan, Ibadan, Nigeria. Rufus Akinyemi28,29, Gennady V. Roshchupkin30,31, Naomi Habib32, Yon Ho Jee33, Jesper Qvist Thomassen34, Vida Abedi35,36, Jara Cárcel-Márquez37,38, Marianne Nygaard39,40, The China Kadoorie Biobank Collaborative Group Hampton L. Leonard41,42,43, Chaojie Yang44,45, Ekaterina Yonova-Doing46,47, Maria J. Knol30, Adam J. Lewis48, Renae L. Judy49, Tetsuro Ago50, Philippe Amouyel51,52,53, Nicole D. Armstrong54, Mark K. Bakker55, Traci M. Bartz56,57, David A. Bennett58, Kuang Lin18, Zhengming Chen18,63, Liming Li101, Iona Y. Millwood18,63, Robin G. Walters18,63, Joshua C. Bis56, Constance Bordes1, Sigrid Børte20,59,60, Anael Cain32, Paul M. Ridker61,62, Junshi Chen372, Robert Clarke18, Rory Collins18, Yu Guo373, Chen Wang374,375, Jun Lv101, Kelly Cho7, Zhengming Chen18,63, Carlos Cruchaga64,65, John W. Cole66,67, Phil L. de Jager13,68, Richard Peto18, Yiping Chen18,63, Zammy Fairhurst-Hunter18, Michael Hill18,63, Rafael de Cid69, Matthias Endres70,71,72,73, Leslie E. Ferreira74, Mirjam I. Geerlings75, Alfred Pozarickij18, Dan Schmidt18, Becky Stevens18, Iain Turnbull18 & Canqing Yu101 Natalie C. Gasca57, Vilmundur Gudnason76,77, Jun Hata78, Jing He48, Alicia K. Heath79, Yuk-Lam Ho7, Aki S. Havulinna80,81, Jemma C. Hopewell82, Hyacinth I. Hyacinth83, 372 China National Center For Food Safety Risk Assessment, Beijing, China. 373Fuwai Hospital, Michael Inouye46,84,85,86,87, Mina A. Jacob88, Christina E. Jeon89, Christina Jern90,91, Chinese Academy of Medical Sciences, National Center for Cardiovascular Diseases, Beijing, Masahiro Kamouchi92, Keith L. Keene93, Takanari Kitazono50, Steven J. Kittner67,94, China. 374Chinese Academy of Medical Sciences, Beijing, China. 375Peking Union Medical Takahiro Konuma6, Amit Kumar24, Paul Lacaze95, Lenore J. Launer96, Keon-Joo Lee97, College, Beijing, China. Kaido Lepik4,98,99,100, Jiang Li35, Liming Li101, Ani Manichaikul44, Hugh S. Markus102, Nicholas A. Marston8,9, Thomas Meitinger103,104, Braxton D. Mitchell105,106, Felipe A. Montellano107,108, Takayuki Morisaki10, Thomas H. Mosley109, Mike A. Nalls41,42,43, The VA Million Veteran Program Børge G. Nordestgaard110,111, Martin J. O’Donnell112, Yukinori Okada6,113,114,115,116,117, N. Charlotte Onland-Moret75, Bruce Ovbiagele118, Annette Peters119,120,121, Daniel C. Posner7,418, Kelly Cho7, Yuk-Lam Ho7, Yan V. Sun136,137, Peter W. F. Wilson136,145 & Bruce M. Psaty56,122,123, Stephen S. Rich44, Jonathan Rosand12,13,124, Marc S. Sabatine8,9, Jennifer E. Huffman7,419 Ralph L. Sacco125,126, Danish Saleheen127, Else Charlotte Sandset128,129, Veikko Salomaa80, Muralidharan Sargurupremraj130, Makoto Sasaki3, Claudia L. Satizabal130,131, Carsten O. Schmidt132, Atsushi Shimizu3, Nicholas L. Smith122,133,134, Kelly L. Sloane135, Yoichi Sutoh3, Yan V. Sun136,137, Kozo Tanno3, Steffen Tiedt2, Turgut Tatlisumak138, The International Stroke Genetics Consortium (ISGC) Nuria P. Torres-Aguila37, Hemant K. Tiwari22, David-Alexandre Trégouët1, Stella Trompet139,140, Anil Man Tuladhar88, Anne Tybjærg-Hansen34,111, Marion van Vugt141, Riina Vibo142, Aniket Mishra1,418, Rainer Malik2,418, Quentin Le Grand1, Yunye He11, Marios K. Georgakis2,12,13, Shefali S. Verma143, Kerri L. Wiggins56, Patrik Wennberg144, Daniel Woo83, Ilana Caro1, Livia Parodi12,13, Hee-Joon Bae23, Ganesh Chauhan24, Michael R. Chong25,26, Peter W. F. Wilson136,145, Huichun Xu105, Qiong Yang131,146, Kyungheon Yoon147, Liisa Tomppo27, Rufus Akinyemi28,29, Vida Abedi35,36, Jara Cárcel-Márquez37,38, Maria J. Knol30, Iona Y. Millwood18,63, Christian Gieger148, Toshiharu Ninomiya78, Hans J. Grabe149,150, Mark K. Bakker55, Joshua C. Bis56, Zhengming Chen18,63, Carlos Cruchaga64,65, J. Wouter Jukema140,151,152, Ina L. Rissanen75, Daniel Strbian27, Young Jin Kim147, Pei-Hsin Chen15, John W. Cole66,67, Leslie E. Ferreira74, Jemma C. Hopewell82, Hyacinth I. Hyacinth83, Ernst Mayerhofer12,13, Christopher D. Anderson12,13,124,173, Marguerite R. Irvin54, Christina Jern90,91, Keith L. Keene93, Steven J. Kittner67,94, Paul Lacaze95, Keon-Joo Lee97, Hieab Adams153,154, Sylvia Wassertheil-Smoller155, Kaare Christensen39,40,156, Hugh S. Markus102, Braxton D. Mitchell105,106, Yukinori Okada6,113,114,115,116,117, Stephen S. Rich44, Mohammad A. Ikram30, Tatjana Rundek125,126, Bradford B. Worrall157,158, G. Mark Lathrop159, Jonathan Rosand12,13,124, Ralph L. Sacco125,126, Muralidharan Sargurupremraj130, Moeen Riaz95, Eleanor M. Simonsick160, Janika Kõrv142, Paulo H. C. França74, Ramin Zand161,162, Claudia L. Satizabal130,131, Turgut Tatlisumak138, David-Alexandre Trégouët1, Daniel Woo83, Kameshwar Prasad24, Ruth Frikke-Schmidt34,111, Frank-Erik de Leeuw88, Thomas Liman71,75,163, Iona Y. Millwood18,63, Daniel Strbian27, Ernst Mayerhofer12,13, Christopher D. Anderson12,13,124,173, Karl Georg Haeusler108, Ynte M. Ruigrok55, Peter Ulrich Heuschmann107,164,165, Sylvia Wassertheil-Smoller155, Tatjana Rundek125,126, Bradford B. Worrall157,158, W. T. Longstreth122,166, Keum Ji Jung18,167, Lisa Bastarache48, Guillaume Paré25,26,168,169, Janika Kõrv142, Paulo H. C. França74, Ramin Zand161,162, Kameshwar Prasad24, Frank-Erik Scott M. Damrauer170,171, Daniel I. Chasman61,62, Jerome I. Rotter172, John-Anker Zwart19,20,59, de Leeuw88, Ynte M. Ruigrok55, W. T. Longstreth122,166, Guillaume Paré25,26,168,169, Teemu J. Niiranen16,17,174, Myriam Fornage175,176, Yung-Po Liaw15,177, Sudha Seshadri130,131,178, Daniel I. Chasman61,62, Jerome I. Rotter172, Myriam Fornage175,176, Sudha Seshadri130,131,178, Israel Fernández-Cadenas37, Robin G. Walters18,63, Christian T. Ruff8,9,419, Israel Fernández-Cadenas37, Robin G. Walters18,63, Mayowa O. Owolabi28,179,419, Mayowa O. Owolabi28,179.419, Jennifer E. Huffman7,419, Lili Milani4,419, Yoichiro Kamatani11,419, Yoichiro Kamatani11,419, Martin Dichgans2,180,181,419 & Stephanie Debette1,182,419 Martin Dichgans2,180,181,419 & Stephanie Debette1,182,419 Regeneron Genetics Center The Biobank Japan The ODYSSEY Study Masaru Koido10,11, Takayuki Morisaki10, Yoichiro Kamatani11,419, Akiko Nagai376, Yoishinori Murakami10 & Koichi Matsuda14 Mina A. Jacob88, Anil Man Tuladhar88, Frank-Erik de Leeuw88, Karlijn F. de Laat379, Anouk G. W. van Norden380, Paul L. de Kort381, Sarah E. Vermeer382, Paul J. A. M. Brouwers383, 376 Department of Public Policy, Institute of Medical Sciences, The University of Tokyo, Tokyo, Rob A. R. Gons384, Paul J. Nederkoorn233, Tom den Heijer385, Robert J. van Oostenbrugge234, Japan. Gert W. van Dijk386 & Frank G. W. van Rooij387 HUNT All-In Stroke Follow-up Studies Bendik Slagsvold Winsvold19,20,21, Sigrid Børte20,59,60, Else Charlotte Sandset128,129, Vilas Menon416, Yun Ju Sung64,65, Chengran Yang64,65, Florence Saillour-Glenisson1,314 & John-Anker Zwart19,20,59, Anne H. Aamodt21,388, Anne H. Skogholt20, Ben M. Brumpton20, Simon Gravel417 Cristen J. Willer220,389, Ingrid Heuch19, Knut Hagen388, Lars G. Fritsche390, Linda M. Pedersen19, Maiken E. Gabrielsen20, Hanne Ellekjær388,391, Wei Zhou392,393, Amy E. Martinsen19,20,59, Espen S. Kristoffersen19,394,395, Jonas B. Nielsen20,220,221, Kristian Hveem20,202,203 & Laurent F. Thomas20,396,397,398 EPIC-CVD Ekaterina Yonova-Doing46,47, N. Charlotte Onland-Moret75, Alicia K. Heath79, Christopher D. Anderson12,13,124,173 & Patrik Wennberg144 The SICFAIL Study 379 Department of Neurology, Haga Hospital, The Hague, The Netherlands. 380Department of Felipe A. Montellano107,108, Karl Georg Haeusler108, Peter Ulrich Heuschmann107,164,165, Neurology, Amphia Hospital, Breda, The Netherlands. 381Department of Neurology, Christoph Kleinschnitz399, Stefan Frantz400 & Kathrin Ungethüm107 Elisabeth-Tweesteden Hospital, Tilburg, The Netherlands. 382Department of Neurology, Rijnstate Hospital, Arnhem, The Netherlands. 383Department of Neurology, Medisch spectrum Twente, Enschede, The Netherlands. 384Department of Neurology, Catharina Hospital, The Generacion Study Eindhoven, The Netherlands. 385Department of Neurology, Franciscus Gasthuis & Vlietland, Rotterdam, The Netherlands. 386Department of Neurology, Catharina Wilhelmina Hospital, Nijmegen, The Netherlands. 387Department of Neurology, Medisch Center Leeuwarden, Jara Cárcel-Márquez37,38, Carlos Cruchaga64,65, Rafael de Cid69, Nuria P. Torres-Aguila37, Leeuwarden, The Netherlands. 388Department of Neuromedicine and Movement Science, Elena Muiño37, Cristina Gallego-Fabrega37,401, Natalia Cullell37,274, Miquel Lledós37,402, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology Laia Llucià-Carol37, Tomas Sobrino403, Francisco Campos403, José Castillo403, (NTNU), Trondheim, Norway. 389Department of Human Genetics, University of Michigan, Ann Marimar Freijó404, Juan Francisco Arenillas405, Victor Obach406, José Álvarez-Sabín407, Arbor, MI, USA. 390Center for Statistical Genetics, Department of Biostatistics, University of Carlos A. Molina407, Marc Ribó407, Jordi Jiménez-Conde238, Jaume Roquer238, Michigan, Ann Arbor, MI, USA. 391Stroke Unit, Department of Internal Medicine, St Olavs Lucia Muñoz-Narbona408, Elena Lopez-Cancio409, Mònica Millán408, Rosa Diaz-Navarro410, Hospital, Trondheim University Hospital, Trondheim, Norway. 392Department of Cristòfol Vives-Bauza410, Gemma Serrano-Heras411, Tomás Segura411, Laura Ibañez64, Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, MI, USA. Laura Heitsch183,259, Pilar Delgado275, Rajat Dhar183, Jerzy Krupinski276, 393 Analytic and Translational Genetics Unit, Massachusetts General Hospital, Boston, MA, Raquel Delgado-Mederos401, Luis Prats-Sánchez401, Pol Camps-Renom401, Natalia Blay69, USA. 394Department of General Practice, University of Oslo, Oslo, Norway. 395Department of Lauro Sumoy412, Joan Montaner278, Jin-Moo Lee183, Joan Martí-Fàbregas401 & Neurology, Akershus University Hospital, Lørenskog, Norway. 396BioCore—Bioinformatics Core Israel Fernández-Cadenas37 Facility, Norwegian University of Science and Technology (NTNU), Trondheim, Norway. 397 Clinic of Laboratory Medicine, St Olavs Hospital, Trondheim University Hospital, Trondheim, Norway. 398Department of Clinical and Molecular Medicine, Norwegian University The Copenhagen City Heart Study of Science and Technology (NTNU), Trondheim, Norway. 399Department of Neurology and Center for Translational Neuro- and Behavioral Sciences (C-TNBS), University Hospital Essen, Jesper Qvist Thomassen34, Børge G. Nordestgaard110,111, Anne Tybjærg-Hansen34,111, University Duisburg-Essen, Essen, Germany. 400Department of Medicine I, University Hospital Ruth Frikke-Schmidt34,111, Peter Schnohr413, Gorm B. Jensen413, Marianne Benn34,111, Würzburg, Würzburg, Germany. 401Stroke Unit, Department of Neurology, Biomedical Shoaib Afzal110,111 & Pia R. Kamstrup110 Research Institute Sant Pau, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain. 402Institute for Biomedical Research of Barcelona (IIBB), National Spanish Research Council (CSIC), Barcelona, Spain. 403Clinical Neurosciences Research Laboratory (LINC), Health Research Institute of Santiago de Compostela (IDIS), Santiago de Compostela, Spain. 404Department The SMART Study of Neurology, Biocruces-Bizkaia Health Research Institute, Bilbao, Spain. 405Stroke Unit, Department of Neurology, University Hospital of Valladolid, Valladolid, Spain. 406Department of Neurology, Hospital Clínic de Barcelona, IDIBAPS, Barcelona, Spain. 407Stroke Unit, Marion van Vugt141, Ina L. Rissanen75, Mirjam I. Geerlings75, Jessica van Setten141, Department of Neurology, Hospital Universitari Vall d’Hebron, Barcelona, Spain. Sander W. van der Laan414 & Jet M. J. Vonk75,415 408 Department of Neurosciences, Hospital Germans Trias I Pujol, Universitat Autònoma de Barcelona, Barcelona, Spain. 409Department of Neurology, University Hospital Central de Asturias (HUCA), Oviedo, Spain. 410Department of Neurology, Son Espases University Hospital, Clinical Research Collaboration for Stroke in Korea (CRCS-K) and Korea Biobank Array (KBA) Illes Balears Health Research Institute (IdISBa), Palma, Spain. 411Department of Neurology, Project University Hospital of Albacete, Albacete, Spain. 412High Content Genomics and Bioinformatics Unit, Germans Trias i Pujol Research Institute (IGTP), Badalona, Spain. Hee-Joon Bae23, Keon-Joo Lee97, Kyungheon Yoon147, Young Jin Kim147 & Bong-Jo Kim148 413 The Copenhagen City Heart Study, Copenhagen University Hospital—Bispebjerg and Frederiksberg, Copenhagen, Denmark. 414Central Diagnostics Laboratory, Division Laboratories, Pharmacy, and Biomedical genetics, University Medical Center Utrecht, Utrecht University, Utrecht, The Netherlands. 415Taub Institute for Research on Alzheimer’s Disease and Helsinki Stroke Project the Aging Brain, College of Physicians and Surgeons, Columbia University, New York, NY, USA. 416 The Centre for Translational and Computational Neuroimmunology, Columbia University Liisa Tomppo27, Veikko Salomaa80, Turgut Tatlisumak138, Gerli Sibolt27, Sami Curtze27, Medical Center, New York, NY, USA. 417Department of Human Genetics, McGill University, Marjaana Tiainen27 & Janne Kinnunen27 Montreal, Quebec, Canada. Article Methods We also performed a cross-ancestry meta-regression using MR-MEGA10. Before the meta-analysis using MR-MEGA, we applied All human research was approved by relevant boards and/or institutions the ‘genomic inflation’ correction option to all of the input files, and for each study (Supplementary Table 56) and was conducted according removed variants with extreme effect size values (log[OR] > 5 or to the Declaration of Helsinki. All of the participants provided written log[OR] < −5), MAF < 0.01, imputation quality scores of less than 0.50 informed consent. and effective allele counts (EAC = 2 × number of cases × MAF × imputa- tion quality score) of less than 6. After the meta-analysis, we considered Study design and phenotypes loci to be genome-wide significant for MR-MEGA P < 5 × 10−8 and show- Information on participating studies (discovery and follow-up), study ing nominal association (P < 0.05) in at least one third of studies in any design, and definitions of stroke and stroke subtypes is provided in individual ancestry group (European, East Asian, African American, the Supplementary Information. Population characteristics of indi- Hispanic and South Asian). vidual studies are provided in Supplementary Table 1. Multitrait association study Genotyping, imputation and GWASs To identify additional stroke-risk loci we used MTAG11 in Europeans and Genotyping methods, pre-imputation quality control of genotypes and East Asians, including traits correlated with specific stroke subtypes, imputation methods of individual cohorts (discovery and follow-up) namely CAD for LAS, atrial fibrillation54 for CES, and WMH55 (an MRI are presented in Supplementary Table 2. High-quality samples and marker of cerebral small vessel disease, available in Europeans only) SNPs underwent imputation using mostly Haplotype Reference Con- for SVS. We also ran an MTAG analysis of AS and AIS, including all three sortium (HRC) or 1000 Genomes phase 1 or phase 3 reference panels correlated traits (CAD, atrial fibrillation, WMH (European)). In European and, less often, TOPMed, HapMap or biobank-specific reference panels. individuals, we used summary statistics of published GWAS analyses Individual studies performed a GWAS using logistic regression (or Cox for CAD56, AF54 and WMH55. In East Asians, we used summary statistics regression in some longitudinal population-based cohorts) testing of published GWAS analyses for CAD57 and atrial fibrillation58 (Supple- association of genotypes with five stroke phenotypes (AS, AIS, CES, mentary Information). Associations were retained when the following LAS and SVS) under an additive effect model, adjusting for age, sex, three conditions were verified: (1) MTAG P value for stroke < 5 × 10−8; principal components of population stratification and study-specific (2) P value for stroke < 0.05 in the univariate GWAS; and (3) MTAG covariates when needed (Supplementary Table 2). P value for stroke less than the P value for any of the included traits in The R package EasyQC along with custom harmonization scripts univariate GWASs. were used to perform the quality control of individual GWAS summary results. Marker names and alleles were harmonized across studies. Independent follow-up of GWAS signals Meta-analyses were restricted to autosomal biallelic SNPs from the First, we sought to replicate internally the 42 stroke-risk loci reach- HRC panel. Duplicate markers were removed. Before the meta-analysis, ing genome-wide significance in IVW meta-analyses within individual we removed variants with extreme effect size values (log[OR] > 5 or ancestries, in at least one other ancestry group among the discovery log[OR] < −5), minor allele frequency (MAF) < 0.01, imputation quality samples, considering both nominal replication levels (P < 0.05) and scores of less than 0.50 and effective allele counts (EAC = 2 × number multiple-testing corrected significance at P < 1.19 × 10−3 (0.05/42). Sec- of cases × MAF × imputation quality score) of less than 6. ond, we gathered independent datasets totalling 89,084 AS (including The overall analytical strategy is shown in Extended Data Fig. 1. 85,546 AIS; and 70.0% European, 15.6% African American, 10.1% East We conducted ancestry-specific fixed-effect IVW meta-analyses in Asian, 4.1% Hispanic and 0.1% South Asian) and 1,013,843 controls for European, East Asian, African American, Hispanic and South Asian external replication of associations with AS and AIS (Supplementary populations, followed by cross-ancestry meta-analyses using Tables 1 and 2). These comprised eight biobanks (82,263 cases, 930,988 METAL50. In each meta-analysis we removed variants with hetero- controls) and four hospital-based cohorts (6,821 cases, 82,855 con- geneity P < 1 × 10−6 and variants available in less than one third of trols). We considered both nominal replication levels (P < 0.05) and the total number of cases and less than one third of the total num- multiple-testing corrected significance at P < 8.2 × 10−4 (0.05/60) and ber of contributing studies. We applied the covariate adjusted P < 1.3 × 10−3 (0.05/29) for follow-up of genome-wide significant loci linkage disequilibrium score regression (cov-LDSC) method to from the IVW and the MR-MEGA/MTAG meta-analyses, respectively ancestry-specific GWAS meta-analyses without GC correction to (two-sided P values were used for both discovery and replication analy- test for genomic inflation and to compute robust SNP-heritability ses). We considered stroke-risk loci as high confidence in the case of estimates in admixed populations51. We conducted cross-ancestry significant internal inter-ancestry and/or external replication after GWAS meta-analyses without genomic correction and with cor- accounting for the number of loci tested, nominally significant repli- rection of the linkage-disequilibrium score intercept for genomic cation in both internal and external replication analyses, or evidence inflation observed in individual ancestry-specific GWASs. We con- of involvement in monogenic stroke; intermediate confidence in the ducted separate GWAS analyses of incident AS and AIS (n = 32,903 and case of nominal significance in either internal inter-ancestry or external n = 16,863) in longitudinal population-based cohort studies. For the replication analyses but not both; and low confidence in the absence meta-analysis combining both incident and prevalent stroke studies, of formal replication. a few incident stroke studies were removed because they were already part of a meta-analysis of stroke GWASs used as an input of the overall Gene-based analyses meta-analysis (WHI, Hisayama, REGARDS, JHS). We considered loci We performed gene-based tests of common variant associations using to be genome-wide significant for P < 5 x 10-8. VEGAS28 and MAGMA9. Both VEGAS2 and MAGMA considered variants We applied the conditional and joint analysis approach7 implemented in the gene or within 10 kb on either side of a gene’s transcription site in the Genome-wide Complex Trait Analysis software52 (GCTA-COJO) to compute a gene-based P value. We performed MAGMA tests using to identify potentially independent signals within the same geno­ the default parameters, whereas the VEGAS2 analyses were performed mic region. We performed GCTA-COJO analyses on (1) European using the ‘-top 10’ parameter that tests enrichment of the top 10% var- GWAS meta-analysis summary statistics using HRC imputed data of iants assigned to a gene accounting for the linkage disequilibrium 6,489 French participants from the 3C study as in ref. 53 and (2) East between variants and the total number of variants within a gene. We Asian-ancestry-specific GWAS meta-analysis summary statistics using used 1000 Genomes phase 3 continental reference samples of European, BBJ data as reference (Supplementary Information). East Asian, African, South Asian and South American (for our Hispanic samples) ancestry and to compute the linkage disequilibrium between (P < 0.05 indicated significance)66. MR analyses were performed in R variants for respective ancestry-specific gene-based analyses. We then v.4.1.1 using the Mendelian Randomization package. meta-analysed ancestry-specific gene-based results, using Stouffer’s For all genetic correlation and MR analyses, we set statistical signifi- method for sample-size-weighted combination of P values. Gene-wide cance at Bonferroni-corrected P < 4.17 × 10−3 in Europeans (correcting significance was defined as P < 2.72 × 10−6, correcting for 18,371 auto- for 12 vascular-risk traits) and P <8.33 × 10−3 in East Asians (correcting somal protein-coding genes tested. for 6 vascular-risk traits). Pathway-based analyses Cross-ancestry fine mapping We used the ancestry-specific gene-based association P values gener- Fine-mapping was performed separately for Europeans and East Asians ated using VEGAS2 to perform pathway analyses for individual ances- using susieR v.0.9.112 on all variants within 3 Mb of the lead variant of try groups, testing enrichment of gene-based P values in Biosystems each genomic risk locus (60 loci reached genome-wide significance pathways with VEGAS2Pathway8,15. For each stroke phenotype, we in the IVW meta-analysis). Unrelated individuals from the UK Biobank meta-analysed the ancestry-specific pathway association P values (n = 420,000) and BBJ (n = 170,000) were used as ancestry-matched using Stouffer’s method considering the number of cases in each linkage-disequilibrium reference panels that fulfil the sample size ancestry-specific GWAS; for example, for AS, we considered 73,652, requirement67. After extracting variants present in the linkage disequi- 27,413, 3,961, 1,516 and 3,640 cases in European-, East Asian-, African librium reference panel, the default settings of susieR were used while American-, Hispanic- and South Asian-specific GWAS analyses to com- allowing for a maximum of 10 putative causal variants in each locus. bine the respective ancestry-specific pathway association P values. The fine-mapping results were checked for potential false-positive Pathway-wide significance was defined at P < 5.01 × 10−6 correcting for findings using a diagnostic procedure implemented in SuSiE. In brief, 9,977 Biosystems pathways tested. we compared observed and expected z-scores for each variant at a given locus and removed the variant if the difference between the observed Shared genetic variation and expected z-score was too high after manual inspection. We com- We examined shared genetic variation with 12 vascular risk factors and pared the variants in credible sets of the same loci between Europeans related disease traits in Europeans using summary statistics of GWASs and East Asians. on SBP59, DBP59, BMI and waist-to-hip ratio60, high density lipoprotein To detect putative causal regulatory variants, we conducted an in (HDL) cholesterol61, LDL cholesterol61, triglycerides61, type 2 diabetes62, silico mutagenesis analysis using MENTR (mutation effect prediction WMH volume55, atrial fibrillation54, CAD56 and venous thromboembo- on non-coding RNA transcription; https://github.com/koido/MENTR), a lism32. We extracted sentinel stroke-risk variants (or a proxy (r2 > 0.9)) machine-learning method to precisely predict transcriptional changes that showed genome-wide significant association (P < 5 × 10−8) with induced by causal variants3,68. The in silico mutations predicted to the aforementioned vascular-risk traits. have strong effects are highly concordant with the observed effects of We then systematically examined genetic correlations and poten- known variants in a cell-type-dependent manner. Furthermore, MENTR tially causal associations between vascular-risk traits and risk of stroke does not use population datasets and is therefore less susceptible using linkage-disequilibrium score regression (LDSC) and MR analyses, to linkage-disequilibrium-dependent association signals, enabling with 12 (in Europeans) and 6 (in East Asians) vascular-risk traits. In indi- precise prediction of the effects of causal variants on transcriptional viduals of European ancestry, we used summary statistics of the afore- changes. From 1,274 variants in the credible sets from the European mentioned GWASs32,54–56,59–62. For the analysis in East Asians, we used and East Asian fine-mapping, we searched FANTOM5 promoters and unpublished GWAS analyses for SBP, DBP, LDL and HDL cholesterol, enhancers, obtained by cap analysis of gene expression, within ±100 kb triglycerides and BMI in up to 53,323 participants of the independent from each variant. As a result, we found 37,878 variant–transcript pairs Tohoku Medical Megabank Project (Supplementary Information). comprising 1,270 variants and 2,350 transcripts. We used MENTR with We used cov-LDSC to compute genetic correlations between stroke the pretrained FANTOM5 347 cell/tissue models + LCL models69–72 and and vascular-risk traits, using European and East Asian GWAS summary extracted reliable predictions using the predetermined robust thresh- files and 1000Gp3v5 reference data of respective continental ances- old (absolute in silico mutation effects ≥ 0.1, achieving >90% concord- tries (considering the recommended subset of high-quality HapMap3 ance for predicting effects on expression). SNPs only). For MR analyses, we constructed genetic instruments for each TWAS and PWAS vascular-risk trait based on genome-wide significant associations We performed TWAS using TWAS-Fusion19 to identify genes of which (P < 5 × 10−8) after clumping for linkage disequilibrium at r2 < 0.01 the expression is significantly associated with stroke risk. We restricted (based on European and East Asian 1000 Genomes reference panels). the analysis to tissues considered to be relevant for cerebrovascular We applied two-sample MR analyses in the GIGASTROKE summary disease, and used precomputed functional weights from 21 publicly statistics separately for individuals of European and East Asian ances- available eQTL reference panels from blood (Netherlands Twin Registry; try based on variant associations derived from the aforementioned Young Finns Study)19,20, arterial and heart (GTEx v.7))17 and brain tis- sources. After extraction of the association estimates and harmoniza- sues (GTEx v.7, CommonMind Consortium)17,18. Moreover, we used the tion of their direction-of-effect alleles, we computed MR estimates with newly developed cross-tissue weights generated in GTEx v.8 using fixed-effect IVW analyses63. As a measure of pleiotropy, we assessed sparse canonical correlation analysis (sCCA) across 49 tissues avail- heterogeneity across the MR estimates for each instrument in the IVW able on the TWAS-Fusion website, including gene expression models MR analyses with Cochran’s Q statistic (P < 0.05 was considered to be for the first three canonical vectors (sCCA1–3), which were shown to significant)64. We further applied alternative MR methods that are more capture most of the gene expression signal73. TWAS-Fusion was then robust to the use of pleiotropic instruments: the weighted median used to estimate the TWAS association statistics between predicted estimator enables the use of invalid instruments under the assumption gene expression and stroke by integrating information from expression that at least half of the instruments used in the MR analysis are valid65; reference panels (SNP-expression weights), GWAS summary statistics MR-Egger regression allows for the estimation of an intercept term, (SNP-stroke effect estimates) and linkage disequilibrium reference provides less precise estimates and relies on the assumption that the panels (SNP correlation matrix)19. Transcriptome-wide significant strengths of potential pleiotropic instruments are independent of their genes (eGenes) and the corresponding eQTLs were determined using direct associations with the outcome66. The intercept obtained from Bonferroni correction, based on the average number of features (5005.8 MR-Egger regression was used as a measure of directional pleiotropy genes) tested across all reference panels and correcting for the 5 stroke Article phenotypes (P < 2.0 × 10−6). eGenes were then tested in conditional summarize variant-level P values into gene level and used the genes analysis as implemented using the Fusion software19. To ensure that with FDR < 0.05 in either MAGMA or VEGAS2 as the disease-risk genes. the observed associations did not reflect random correlation between We then used GREP22 to perform a series of Fisher’s exact tests for the gene expression and non-causal variants associated with stroke, we enrichment of the disease-risk genes in the drug-target genes involved performed a colocalization analysis on the conditionally significant in the drug indication categories, Anatomical Therapeutic Chemical genes (P < 0.05) to estimate the posterior probability of a shared causal Classification System codes. variant between the gene expression and trait association (PP4)74. We used a prior probability of P < 2.0 × 10−6 for the stroke association. Genes Negative correlation tests between genetically determined and presenting a PP4 ≥ 0.75, for which eQTLs did not reach genome-wide compound-regulated gene expression profiles. We nominated the significance in association with stroke, and were not in linkage dis- compounds with inverse effects on gene expression against genetically equilibrium (r2 < 0.01) with any of the lead SNPs of genome-wide sig- determined gene expression by using Trans-Phar24. In brief, genetically nificant risk loci for stroke, were considered to be new, i.e. not within determined case–control gene expression was inferred for 44 tissues a genome-wide significant stroke risk locus. in the Genotype-Tissue Expression project (v.7)17 with FOCUS83, and Using similar parameters in TWAS-Fusion19, we also performed the genes in the top decile for the absolute value of the z-score were a proteome-wide association study. For this analysis, we used the used for the following correlation analysis. The Library of Integrated precomputed weights for protein expression in DLPFC75 from the ROS/ Network-based Cellular Signatures project (LINCS) CMAP L1000 library MAP study (n = 376 individuals, n = 1,475 proteins)76 and the Banner data84 were used for the compound library. After matching the tissues Sun Health Institute study (n = 152 individuals, n = 1,145 proteins)77. in GTEx with the cell lines in the LINCS L1000 library, we performed Proteome-wide significant genes and the corresponding pQTLs were a series of Spearman’s rank correlation tests for 308,872 pairs of ge- determined using Bonferroni correction, on the number of proteins netically determined and compound-perturbed tissue- or cell-type tested across the reference panel and correcting for the 5 stroke phe- specific gene expression profiles. We prioritized compounds with notypes (P < 1.7 × 10−4 for ROS/MAP and P < 2.2 × 10−8 for the Banner FDR < 0.1, as we previously showed that the compounds with FDR < 0.1 Sun Health Institute study). We then followed the same method as contained plausible therapeutic targets with literature supports4. described for the TWAS. Endophenotype MR. To pin-point the disease-causing proteins that Brain single-cell expression analyses were targeted by existing drugs, we performed MR analyses (specifi- Single-nucleus RNA-sequencing data of the DLPFC region of 24 ageing cally, a Wald ratio test) by using lead variants in pQTL as instrumental individuals chosen to represent the range of pathologic and clinical variables and five stroke phenotypes as outcomes: AS, AIS, CES, LAS diagnoses of AD dementia, from the ROS/MAP cohorts, was obtained21. and SVS. We used the tier 1 lead variants defined in ref. 85 to avoid con- RNA profiles of cells annotated as endothelial, pericytes or smooth founding by horizontal pleiotropy. The tier 1 variants, summarized muscle cells and vascular leptomeningeal cells (VLMC) were used, and from five pQTL studies (n = 997 to 6,861)86–90, did not include variants a pseudobulk RNA profile was generated for each cell type by averaging with heterogeneous effect sizes among the studies or with a number of the expression of all genes across the cells. Average expression levels and associated proteins of larger than five. We restricted the lead variants to the percentage of expressed genes were calculated for genes of interest the variants associated with drug-target proteins. For the lead variants using the DotPlot function from the Seurat package v.4.0.4 in R v.4.1.1. of pQTLs that were missing in the stroke GWAS summary statistics, the We also conducted a cell-type enrichment analysis using the STEAP proxy variants with the largest r2 were used if the r2 was greater than 0.8 pipeline (https://github.com/ComPopBio/STEAP). This is an extension (1000 Genomes, European). In total, we used 277 lead variants for 218 of CELLECT and uses S-LDSC78, MAGMA9 and H-MAGMA79 for enrich- drug-target proteins for MR and considered FDR < 0.05 as the thresh- ment analysis. Stroke GWAS summary statistics were first munged. old to identify significant associations. We used the TwoSampleMR R Expression specificity profiles were then calculated using human package91 for MR analysis. As post-MR quality controls, we performed and mouse single-cell RNA-seq databases (Supplementary Table 28). (1) a directionality check of causal relationships by Steiger filtering92 Cell-type enrichment was calculated using three models: MAGMA, and (2) colocalization analysis for the proteins with FDR < 0.05. To H-MAGMA (incorporating chromatin interaction profiles from human examine colocalization assuming multiple causal variants per locus, brain tissues in MAGMA) and stratified linkage-disequilibrium score coloc74 was applied to the decomposed signals by SuSiE12 for the variants regression. P values were corrected for the number of independent within 500 kb upstream and downstream of the lead variants (coloc + cell types in each database (Bonferroni correction). SuSiE)93. If SuSiE did not converge after 10,000 iterations, coloc was used instead. coloc + SuSiE and coloc were run with their respective Genomics-driven drug discovery default parameters. For the two pQTL studies without public summary We used three methodologies for in-depth genomics-driven drug dis- statistics86,90, we compared the r2 between the lead variants of the pQTL covery as described previously4: (1) an overlap enrichment analysis of study and the stroke GWAS. We considered that colocalization occurred disease-risk genes in drug-target genes in medication categories; (2) when the maximum posterior probability (that is, PP.H4) was greater negative correlation tests between genetically determined case–con- than 0.75 or r2 was greater than 0.8. trol gene expression profiles and compound-regulated gene expres- To provide further support for our findings, we conducted MR analy- sion profiles; and (3) endophenotype MR. Details of the methods are ses with two additional recent independent pQTL datasets, using the described in the following sections. For the overlap enrichment analysis same methodology and significance thresholds (FDR < 0.05 for MR and and the endophenotype MR-nominated drug targets, we curated drug PP.H4 > 0.75 for colocalization) as above: one study comprised both candidates from four major drug databases: DrugBank23, Therapeutic plasma (n = 529) and cerebrospinal fluid (n = 835) pQTL datasets94, the Target Database (TTD)80, PharmGKB81 and Open Target Platform82. As second is one of the largest plasma pQTL studies conducted in 35,559 for the endophenotype MR, we curated drugs with opposite effects Icelandic individuals95. against the signs of the MR effect estimates. By contrast, the negative correlation tests directly prioritized candidate compounds. We manu- Protective rare variants ally curated supporting evidence for candidate drugs and c­om­po­un­ds. For the five genes targeted by inhibitors—VCAM1, F11, KLKB1, LAMC2 and GP1BA—we extracted the associations of rare deleterious variants Overlap enrichment analysis of disease-risk genes in drug-target (MAF < 0.01) with stroke and stroke-related traits from the gene-based genes in medication categories. We ran MAGMA9 and VEGAS28 to burden tests in the whole-exome sequencing data of >450,000 UK Biobank participants31. As stroke and stroke-related traits, we extracted from the GIGASTROKE project, and 12 GWAS analyses of vascular-risk 30 traits belonging to 9 vascular risk factor and disease categories traits from other groups (Extended Data Fig. 10). To avoid the overlap (Supplementary Table 35). We applied Bonferroni correction and the of participants across datasets, the GWAS summary statistics for stroke corrected P-value threshold was 0.05/5/30 = 3.33 × 10−4 (5 and 30 rep- outcomes were recalculated for the iPGS analysis by excluding the EstBB resent the number of tested genes and traits, respectively). from the meta-analysis of GIGASTROKE studies. To enable compari- son with a previous European iPGS model based on the MEGASTROKE PheWAS GWAS5, we incorporated 12 GWAS analyses of vascular-risk traits (atrial PheWAS analysis was performed using R (v.4.0.3). We used the PheWAS fibrillation, CAD, T2D, SBP, DBP, TC, LDL-C, HDL-C, TG, BMI, height and R package96 (https://github.com/PheWAS/PheWAS) function create- smoking)54,56,59–61,107,108 into the GIGASTROKE-based iPGS model. The Phenotypes to translate ICD10 diagnosis codes into phecodes for the iPGS model for Europeans was further evaluated in two external cohorts PheWAS analysis. We tested the associations between phecodes and of European ancestry (MVP and pooled data of clinical trials) as well as genetic variants using logistic regression and adjusting for sex, birth in two studies of participants with African ancestry (MVP and SIREN). year and ten genotype PCs. We applied Bonferroni correction to select For the East Asian iPGS model, we used BBJ data for the model train- statistically significant associations (number of tested phecodes: 1,809; ing and evaluation. The model-training dataset was composed of 577 number of tested SNPs: 8; corrected P-value threshold: 0.05/(1,809 × 8) = cases of AIS and 9,232 control individuals, whereas there were 1,470 3.45 × 10−6). The results were visualized using the PheWAS library. cases of AIS and 40,459 control individuals in the model-evaluation To further characterize the associations of the genetic variants with dataset. The mean ± s.d. of age at recruitment was 69.2 ± 10.8 years other phenotypes, we searched for all eight SNPs in the PhenoScanner for cases and 66.5 ± 12.5 years for controls in the model evaluation database97,98. dataset. The percentage of male participants was 70.0% for cases and 53.1% for controls. The two case–control datasets were not included in Polygenic risk prediction the meta-analysis of GIGASTROKE studies and, therefore, the overlap We constructed iPGS models for stroke in European and East Asian of participants across datasets was avoided. To derive the East Asian individuals (Extended Data Fig. 10). For each ancestry, independent iPGS model, we incorporated 5 ancestry-specific and 5 cross-ancestry datasets were used for model training and evaluation. We used as input stroke GWAS analyses (AS, AIS, LAS, SVS and CES) from the GIGASTROKE summary statistics data of multiple GWAS analyses for stroke outcomes project, and 12 GWAS analyses of vascular-risk traits (Extended Data and vascular-risk traits to derive iPGS models. We denote the number of Fig. 10). The iPGS model for East Asian individuals was further evalu- input GWASs as N. For each of the N GWAS summary data, 37 candidate ated in an external study of East Asian ancestry (TWB). single-trait polygenic score (sPGS) models were generated using the P+T99,100, LDpred101 and PRScs102 algorithms with an ancestry-specific GRS in clinical trials linkage-disequilibrium reference panel from the 1000 Genomes Pro- Participants who had consented for genetic testing and who were of ject103 (Supplementary Methods). The plink (v.1.90b6.8)104, LDpred European ancestry from the ENGAGE AF-TIMI 48 (effective anticoagula- (v.1.0.11)101 and PRScs.py (5 June 2021)102 programs were used to com- tion with factor Xa next generation in atrial fibrillation)109, SOLID-TIMI 52 pute the P+T, LDpred and PRScs models, respectively. Subsequently, (stabilization of plaques using darapladib)110, SAVOR-TIMI 53 (saxaglip- among the 37 candidate models, the best sPGS model, which was tin assessment of vascular outcomes recorded in patients with diabetes defined as the model that showed a maximal improvement in AUC mellitus)111, PEGASUS-TIMI 54 (prevention of cardiovascular events in over a base model (age, sex and top five PCs were included in the base patients with prior heart attack using ticagrelor compared to placebo model), was selected using the model training dataset5,100. Then, N best on a background of aspirin)112 and FOURIER (further cardiovascular sPGS models were selected from the N input GWASs. Among the N best outcomes research with PCSK9 inhibition in patients with elevated sPGS models, we retained models that were significantly associated risk)113 trials were included in this analysis. Methods for genotyping and with AIS in the model-training dataset (Bonferroni-corrected P < 0.05). imputation have previously been published35,114 and are summarized Then, each retained best sPGS was z-transformed (zero mean and unit in Supplementary Table 2. A set of 58 sentinel variants at stroke-risk s.d.) over the model-training dataset, followed by elastic-net logistic loci identified in the IVW meta-analysis was used to calculate a GRS for regression105 to model the associations between the N sPGS variables each trial participant and identify tertiles of genetic risk (Supplemen- and AIS with the adjustments for age, sex and top five genetic PCs. tary Table 57). A Cox model was used to estimate HRs for ischaemic Two regularization parameters (α and λ) were optimized using tenfold stroke associated with the quantitative GRS and across genetic risk cross-validation. Coefficients (weights) for the retained sPGS models groups, adjusted for clinical risk factors (age, sex, hypertension, hyper- were then determined by elastic-net logistic regression with the optimal lipidaemia, diabetes, smoking, CAD, atrial fibrillation and congestive regularization parameters, followed by integration of the sPGS models heart failure) and the first five principal components of population into a single iPGS model according to a formula presented previously5. stratification. Analyses were conducted primarily in participants of Elastic-net regression was performed using the glmnet R package106. European ancestry (n = 51,288, with 960 incident AIS)—with secondary The predictive ability of the iPGS model was estimated using the analyses in the much smaller East Asian (n = 1,312, with 27 incident AIS) model-evaluation dataset, whereby we evaluated the improvement ancestry subset—using the AS cross-ancestry IVW meta-analysis effect in C-index for a prospective cohort dataset or AUC for a case-control estimates as weights for the primary analysis and ancestry-specific, dataset over a base model that includes age, sex and top five genetic PCs. as well as AIS effect estimates for secondary analyses. We also looked We used EstBB data for the model training and evaluation of iPGS separately at associations with incident stroke in participants with and model in Europeans. The model-training dataset was composed of 1,003 without previous stroke. cases of prevalent AIS at the baseline and 8,997 control individuals. The control individuals were randomly selected among EstBB participants Reporting summary who had no history of AS at the baseline and who did not develop AS dur- Further information on research design is available in the Nature ing the follow-up. The remaining 102,099 EstBB participants were used Research Reporting Summary linked to this article. for the model evaluation (mean ± s.d. age at the baseline, 44.0 ± 15.7 years; 37.8% men). Among the participants in the model-evaluation dataset, 1,128 cases of incident AIS were observed during 4.6 ± 4.8 years. Data availability To derive the European iPGS model, we incorporated 5 ancestry-specific Summary statistics generated by the GIGASTROKE consortium across and 5 cross-ancestry stroke GWAS analyses (AS, AIS, LAS, SVS and CES) ancestries and stroke subtypes are available in the GWAS Catalog Article (GCST90104534–GCST90104563). The integrated polygenic risk 69. Andersson, R. et al. An atlas of active enhancers across human cell types and tissues. Nature 507, 455–461 (2014). score models of stroke in Europeans and East Asians are available in 70. Forrest, A. R. et al. A promoter-level mammalian expression atlas. Nature 507, 462–470 the PGS Catalog (PGS002724 and PGS002725). Individual level data (2014). can be requested directly from the authors of the contributing stud- 71. Hon, C. C. et al. An atlas of human long non-coding RNAs with accurate 5′ ends. Nature 543, 199–204 (2017). ies, listed in Supplementary Table 1. Single-nucleus RNA-seq data have 72. Garieri, M. et al. The effect of genetic variation on promoter usage and enhancer activity. been deposited in the SYNAPSE database as part of the Religious Orders Nat. Commun. 8, 1358 (2017). Study and Memory and Aging Project (ROSMAP) (https://www.syn- 73. Feng, H. et al. Leveraging expression from multiple tissues using sparse canonical correlation analysis and aggregate tests improves the power of transcriptome-wide apse.org) and through the RADC Resource Sharing Hub (https://www. association studies. PLoS Genet. 17, e1008973 (2021). radc.rush.edu). We used publicly available data from GTEx (https:// 74. Giambartolomei, C. et al. Bayesian test for colocalisation between pairs of genetic gtexportal.org/home/), the Gusev laboratory (http://gusevlab.org/ association studies using summary statistics. PLoS Genet. 10, e1004383 (2014). 75. Wingo, A. P. et al. Integrating human brain proteomes with genome-wide association data projects/fusion/), the FinnGen Freeze 7 cohort (https://www.finngen. implicates new proteins in Alzheimer’s disease pathogenesis. Nat. Genet. 53, 143–146 fi/en/access_results), PhenoScanner v.2 database (http://www.phenos- (2021). canner.medschl.cam.ac.uk), pQTL summary statistics (https://doi. 76. Bennett, D. A. et al. Religious orders study and rush memory and aging project. J. Alzheimers Dis. 64, S161–S189 (2018). org/10.1038/s41588-020-0682-6, http://www.phpc.cam.ac.uk/ceu/ 77. Beach, T. G. et al. Arizona study of aging and neurodegenerative disorders and brain and proteins/, http://metabolomics.helmholtz-muenchen.de/pgwas/index. body donation program. Neuropathology 35, 354–389 (2015). php, https://zenodo.org/record/264128), deCODE genetics (https:// 78. Finucane, H. K. et al. Partitioning heritability by functional annotation using genome-wide association summary statistics. Nat. Genet. 47, 1228–1235 (2015). www.decode.com/summarydata/) and summary statistics using the 79. Sey, N. Y. A. et al. A computational tool (H-MAGMA) for improved prediction of UK Biobank whole-exome sequencing (https://doi.org/10.1038/s41586- brain-disorder risk genes by incorporating brain chromatin interaction profiles. Nat. 021-04103-z). Neurosci. 23, 583–593 (2020). 80. Chen, X., Ji, Z. L. & Chen, Y. Z. TTD: Therapeutic Target Database. Nucleic Acids Res. 30, 412–415 (2002). 81. Whirl-Carrillo, M. et al. An evidence-based framework for evaluating pharmacogenomics Code availability knowledge for personalized medicine. Clin. Pharmacol. Ther. 110, 563–572 (2021). 82. Ochoa, D. et al. Open Targets Platform: supporting systematic drug-target identification The code for computation of the integrated polygenic risk score and prioritisation. Nucleic Acids Res. 49, D1302–D1310 (2021). of stroke are available at GitHub (https://github.com/hacchy1983/ 83. Mancuso, N. et al. Probabilistic fine-mapping of transcriptome-wide association studies. Nat. Genet. 51, 675–682 (2019). iPGS-construction). The drug discovery analysis was conducted using 84. Subramanian, A. et al. A next generation connectivity map: L1000 platform and the first the following publicly available tools: GREP (https://github.com/saori- 1,000,000 profiles. Cell 171, 1437–1452 (2017). sakaue/GREP), Trans-Phar (https://github.com/konumat/Trans-Phar), 85. Zheng, J. et al. Phenome-wide Mendelian randomization mapping the influence of the plasma proteome on complex diseases. Nat. Genet. 52, 1122–1131 (2020). and the TwoSampleMR (https://mrcieu.github.io/TwoSampleMR/), 86. Emilsson, V. et al. Co-regulatory networks of human serum proteins link genetics to coloc (https://chr1swallace.github.io/coloc/) and susieR (https://ste- disease. Science 361, 769–773 (2018). phenslab.github.io/susieR/index.html) R packages. 87. Folkersen, L. et al. Mapping of 79 loci for 83 plasma protein biomarkers in cardiovascular disease. PLoS Genet. 13, e1006706 (2017). 88. Suhre, K. et al. Connecting genetic risk to disease end points through the human blood 50. Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis of genomewide plasma proteome. Nat. Commun. 8, 14357 (2017). association scans. Bioinformatics 26, 2190–2191 (2010). 89. Sun, B. B. et al. Genomic atlas of the human plasma proteome. Nature 558, 73–79 (2018). 51. Luo, Y. et al. Estimating heritability and its enrichment in tissue-specific gene sets in 90. Yao, C. et al. Genome-wide mapping of plasma protein QTLs identifies putatively causal admixed populations. Hum. Mol. Genet. 30, 1521–1534 (2021). genes and pathways for cardiovascular disease. Nat. Commun. 9, 3268 (2018). 52. Yang, J., Lee, S. H., Goddard, M. E. & Visscher, P. M. GCTA: a tool for genome-wide 91. Hemani, G. et al. The MR-Base platform supports systematic causal inference across the complex trait analysis. Am. J. Hum. Genet. 88, 76–82 (2011). human phenome. eLife 7, e34408 (2018). 53. 3C Study Group. Vascular factors and risk of dementia: design of the Three-City Study 92. Hemani, G., Tilling, K. & Davey Smith, G. Orienting the causal relationship between and baseline characteristics of the study population. Neuroepidemiology 22, 316–325 imprecisely measured traits using GWAS summary data. PLoS Genet. 13, e1007081 (2017). (2003). 93. Wallace, C. A more accurate method for colocalisation analysis allowing for multiple 54. Nielsen, J. B. et al. Biobank-driven genomic discovery yields new insight into atrial causal variants. PLoS Genet. 17, e1009440 (2021). fibrillation biology. Nat. Genet. 50, 1234–1239 (2018). 94. Yang, C. et al. Genomic atlas of the proteome from brain, CSF and plasma prioritizes 55. Sargurupremraj, M. et al. Cerebral small vessel disease genomics and its implications proteins implicated in neurological disorders. Nat. Neurosci. 24, 1302–1312 (2021). across the lifespan. Nat. Commun. 11, 6285 (2020). 95. Ferkingstad, E. et al. Large-scale integration of the plasma proteome with genetics and 56. Nikpay, M. et al. A comprehensive 1,000 Genomes-based genome-wide association disease. Nat. Genet. 53, 1712–1721 (2021). meta-analysis of coronary artery disease. Nat. Genet. 47, 1121–1130 (2015). 96. Carroll, R. J., Bastarache, L. & Denny, J. C. R PheWAS: data analysis and plotting tools for 57. Ishigaki, K. et al. Large-scale genome-wide association study in a Japanese population phenome-wide association studies in the R environment. Bioinformatics 30, 2375–2376 identifies novel susceptibility loci across different diseases. Nat. Genet. 52, 669–679 (2014). (2020). 97. Staley, J. R. et al. PhenoScanner: a database of human genotype-phenotype associations. 58. Low, S. K. et al. Identification of six new genetic loci associated with atrial fibrillation in Bioinformatics 32, 3207–3209 (2016). the Japanese population. Nat. Genet. 49, 953–958 (2017). 98. Kamat, M. A. et al. PhenoScanner V2: an expanded tool for searching human 59. Evangelou, E. et al. Genetic analysis of over 1 million people identifies 535 new loci genotype-phenotype associations. Bioinformatics 35, 4851–4853 (2019). associated with blood pressure traits. Nat. Genet. 50, 1412–1425 (2018). 99. Purcell, S. M. et al. Common polygenic variation contributes to risk of schizophrenia and 60. Pulit, S. L. et al. Meta-analysis of genome-wide association studies for body fat bipolar disorder. Nature 460, 748–752 (2009). distribution in 694 649 individuals of European ancestry. Hum. Mol. Genet. 28, 100. Khera, A. V. et al. Genome-wide polygenic scores for common diseases identify 166–174 (2019). individuals with risk equivalent to monogenic mutations. Nat. Genet. 50, 1219–1224 61. Willer, C. J. et al. Discovery and refinement of loci associated with lipid levels. Nat. Genet. (2018). 45, 1274–1283 (2013). 101. Vilhjálmsson, B. J. et al. Modeling linkage disequilibrium increases accuracy of polygenic 62. Xue, A. et al. Genome-wide association analyses identify 143 risk variants and putative risk scores. Am. J. Hum. Genet. 97, 576–592 (2015). regulatory mechanisms for type 2 diabetes. Nat. Commun. 9, 2941 (2018). 102. Ge, T., Chen, C.-Y., Ni, Y., Feng, Y.-C. A. & Smoller, J. W. Polygenic prediction via Bayesian 63. Burgess, S., Butterworth, A. & Thompson, S. G. Mendelian randomization analysis with regression and continuous shrinkage priors. Nat. Commun. 10, 1776 (2019). multiple genetic variants using summarized data. Genet. Epidemiol. 37, 658–665 (2013). 103. The 1000 Genomes Project Consortium. A global reference for human genetic variation. 64. Bowden, J., Hemani, G. & Davey Smith, G. Invited Commentary: detecting individual and Nature 526, 68–74 (2015). global horizontal pleiotropy in Mendelian randomization—a job for the humble 104. Chang, C. C. et al. Second-generation PLINK: rising to the challenge of larger and richer heterogeneity statistic? Am. J. Epidemiol. 187, 2681–2685 (2018). datasets. Gigascience 4, 7 (2015). 65. Hartwig, F. P., Davey Smith, G. & Bowden, J. Robust inference in summary data Mendelian 105. Zou, H. & Hastie, T. Regularization and variable selection via the elastic net. J. R. Stat. Soc. randomization via the zero modal pleiotropy assumption. Int. J. Epidemiol. 46, 1985–1998 B 67, 301–320 (2005). (2017). 106. Friedman, J., Hastie, T. & Tibshirani, R. Regularization paths for generalized linear models 66. Bowden, J., Davey Smith, G. & Burgess, S. Mendelian randomization with invalid via coordinate descent. J. Stat. Softw. 33, 1–22 (2010). instruments: effect estimation and bias detection through Egger regression. Int. J. 107. Mahajan, A. et al. Fine-mapping type 2 diabetes loci to single-variant resolution using Epidemiol. 44, 512–525 (2015). high-density imputation and islet-specific epigenome maps. Nat. Genet. 50, 1505–1513 67. Benner, C. et al. Prospects of fine-mapping trait-associated genomic regions by using (2018). summary statistics from genome-wide association studies. Am. J. Hum. Genet. 101, 108. Tobacco and Genetics Consortium. Genome-wide meta-analyses identify multiple loci 539–551 (2017). associated with smoking behavior. Nat. Genet. 42, 441–447 (2010). 68. Zhou, J. et al. Deep learning sequence-based ab initio prediction of variant effects on 109. Giugliano, R. P. et al. Edoxaban versus warfarin in patients with atrial fibrillation. N. Engl. J. expression and disease risk. Nat. Genet. 50, 1171–1179 (2018). Med. 369, 2093–2104 (2013). 110. O’Donoghue, M. L. et al. Effect of darapladib on major coronary events after an acute has consultancies with Anthos, Bayer, Bristol Myers Squibb, Boehringer Ingelheim, Daiichi coronary syndrome: the SOLID-TIMI 52 randomized clinical trial. JAMA 312, 1006–1015 Sankyo, Janssen and Pfizer, institutional research grant to the TIMI Study Group at Brigham and (2014). Women’s Hospital from Anthos, AstraZeneca, Boehringer Ingelheim, Daiichi Sankyo, Janssen, 111. Scirica, B. M. et al. Saxagliptin and cardiovascular outcomes in patients with type 2 National Institutes of Health and Novartis, and consultancies with Anthos, Bayer, Bristol Myers diabetes mellitus. N. Engl. J. Med. 369, 1317–1326 (2013). Squibb, Boehringer Ingelheim, Daiichi Sankyo, Janssen and Pfizer. T.H. receives personal fees 112. Bonaca, M. P. et al. Long-term use of ticagrelor in patients with prior myocardial from Genome Analytics Japan; J.C.H. is supported by a personal fellowship from the British infarction. N. Engl. J. Med. 372, 1791–1800 (2015). Heart Foundation (FS/14/55/30806), and acknowledges additional support from the Nuffield 113. Sabatine, M. S. et al. Evolocumab and clinical outcomes in patients with cardiovascular Department of Population Health (NDPH), University of Oxford, the British Heart Foundation disease. N. Engl. J. Med. 376, 1713–1722 (2017). Centre for Research Excellence, Oxford, and the Oxford Biomedical Research Centre. J.C.H. 114. Marston, N. A. et al. Predicting benefit from evolocumab therapy in patients with holds steering committee and Data and Safety Monitoring Board (DSMB) positions for various atherosclerotic disease using a genetic risk score: results from the FOURIER trial. cardiovascular randomized controlled trials, and is a principal investigator/co-principal Circulation 141, 616–623 (2020). investigator of research grants from industry related to cardiovascular clinical trials and observational studies that are governed by University of Oxford contracts that protect personal independence. NDP.H also has a staff policy of not taking personal payments from Acknowledgements Detailed acknowledgements are provided in the Supplementary industry (further details can be found online; https://www.ndph.ox.ac.uk/files/about/ndph- Information. We thank the participants and staff of contributing studies. We also thank Michèle independence-of-research-policy-jun-20.pdf/@@download); S.S. has consultancies with M. Sale, who passed away in 2018, for her important contributions to some of the studies Biogen; P.U.H. reports grants from German Ministry of Research and Education, during the included in this manuscript, as principal investigator of one of the grants that funded “The Sea conduct of the study, research grants from the German Ministry of Research and Education, Islands Genetics Network (SIGNET)” (R01 DK084350). She contributed specifically to European Union, Charité–Universitätsmedizin Berlin, Berlin Chamber of Physicians, German SIGNET-REGARDS from the REasons for Geographic And Racial Differences in Stroke Parkinson Society, University Hospital Würzburg, Robert Koch Institute, German Heart (REGARDS, U01 NS041588) and served as a key contributor to SiGN, COMPASS and Foundation, Federal Joint Committee (G-BA) within the Innovationfond, German Research METASTROKE. Foundation, Bavarian State (ministry for science and the arts), German Cancer Aid, Charité– Universitätsmedizin Berlin (within Mondafis; supported by an unrestricted research grant to Author contributions S. Debette, M.D., Y.K., L.M., J.E.H., M.O.O. and C.T.F. jointly supervised the Charité from Bayer), University Göttingen (within FIND-AF randomized; supported by an the research. S. Debette and M.D. designed and conceived the study. D.C., M.F., M.N., S.N., unrestricted research grant to the University Göttingen from Boehringer- Ingelheim), University T. Konuma, Y.O., J.Q.T., R.F.-S., S. Trompet, J.B., T.B., K.W., M.R., Y.-H.J., B.W., S.B., H.L., M.A.N., C.Y., Hospital Heidelberg (within RASUNOA-prime; supported by an unrestricted research grant A. Mishra, S.R., J.I.R., M.C., F.K., T.H., Y.S., A.S., G.C., A.K., D. Strbian, Q.Y., F.V., J.L., A.C., N.H., T.J., to the University Hospital Heidelberg from Bayer, BMS, Boehringer-Ingelheim and Daiichi K.K., K. Lepik, J.C.-M., N.P.T.-A., R.M., M.G., J.E.H., E.Y.-D., M. Shi, Y.H., M. Koido, A. Mishra, Q.L.-G., Sankyo), outside the submitted work; K.G.H. reports a study grant by Bayer, lecture fees/ I.C., M.V.V., R.W., K. Lin, M.J.K., A.L., D.P., G.-V.R., H.-J.B., H.T., J.E.H., J. He, K.-J.L., L.T., L.B., advisory board fees from Abbott, Alexion, AMARIN, AstraZeneca, Bayer, Biotronik, Boehringer V. Srinivasasainagendra, Y.J.K. and Y.-C.L. contributed to bioinformatics analyses. Y.R., M.B., C.A., Ingelheim, Bristol-Myers-Squibb, Daiichi Sankyo, Edwards Lifesciences, Medtronic, Pfizer, D.W., P.R., T. Meitinger, K. Cho, K. Christensen, Yi-Ching Liaw, Yung-Po Liaw, B.N., A.T.-H., R.F.-S., Premier Research, SUN Pharma and W. L. Gore & Associates; H.J.G. has received travel grants J.W.J., M.E., T.B., K.W., M.J., F.-E.d.L., P.L., M.R., L.F., P.F., C.J., K.L.K., H.H., Y.-H.J., C.E.J., J.K., R.V., and speakers honoraria from Fresenius Medical Care, Neuraxpharm, Servier and Janssen Cilag B.W., S.B., E.C.S., J.-A.Z., H.S.M., N.G., M.C., G.P., M.O’D., N.M., F.K., M. Sasaki, C.R., K.T., M.S.S., as well as research funding from Fresenius Medical Care; J.M.M.H. is full time employee of K.P., D. Strbian, J.C.H., S.S., A.H., L.L., V.G., N.S., D.-A.T., R.S., T.R., H.A., M.A.I., P.H., K.-G.H., F.M., Novo Nordisk; E.Y.-D. is full-time employee of Novo Nordisk. S. Damrauer receives research V.A., R.Z., S.W.-S., M.A.I., S.K., B.M., H.X., J.C., C.-O.S., L.M., J.R., D. Saleheen, R.D.C., J. Hata, support from RenalytixAI and personal consulting fees from Calico Labs, outside the scope of J.M.M.H., T.N., T.A., M. Koido, T. Kitazono, S. Tiedt, M.D., C.G., A.P., T. Morisaki, T. Meitinger, the current research. H.B. reports grants from AstraZeneca, AstraZeneca Korea, Bayer Korea, M. Kamouchi, Y.K., S. Debette, I.R., M.I., N.A., I.M., A.L., A.K.H., C.C., D.P., D.B., J. He, K.S., L.T., Boehringer Ingelheim Korea, Boryung Pharmaceutical, Bristol Myers Squibb, Bristol Myers L.B., M.O., M.I.G., N.C.O.-M., P.W., P.D.-J., R.J., R.A., S. Damrauer, S.V., T.T., V. Salomaa, Y.-L.H., Squibb Korea, Chong Gun Dang Pharmaceutical, Daiichi Sankyo, Daiichi Sankyo Korea, Dong-A P.-H.C., K. Lee and Y.-P.L. contributed samples and phenotyping. S. Debette, M.D., Y.K., L.M., ST, Esai, Jeil Pharmaceutical, JLK, Korean Drug, SAMJIN Pharm., Servier Korea, Shinpoong J.E.H., M.O.O., C.T.F., A. Mishra, R.M., T.H., T.J., Y.H., M. Koido, M. Shi, K.K. and S.N. wrote and Pharm., Shire International and Yuhan Corporation, and personal fees from Amgen Korea, Esai edited the manuscript. All of the authors provided critical revision. Korea, Otsuka Korea, Takeda Korea and Viatris Korea outside the submitted work. C.C. has received research support from GSK. The funders of the study had no role in the collection, Competing interests C.D.A. has received sponsored research support from Bayer, and has analysis or interpretation of data, in the writing of the report or in the decision to submit the consulted for ApoPharma; T. Konuma is an employee of JAPAN TOBACCO; M. E. reports grants paper for publication. C.C. is a member of the advisory board of Vivid genetics; F.A.M. is from Bayer and fees paid to the Charité from Abbot, AstraZeneca, Bayer, Boehringer Ingelheim, supported by the German Research Foundation (Deutsche Forschungsgemeinschaft (DFG) BMS, Daiichi Sankyo, Amgen, GSK, Sanofi, Covidien, Novartis and Pfizer, all outside the within the UNION-CVD Clinician-Scientist Programme (project number 413657723) and has submitted work; B.M.P. serves on the steering committee of the Yale Open Data Access Project been previously supported by a MD/PhD Fellowship of the Interdisciplinary Center for Clinical funded by Johnson & Johnson; P.A. works with Fondation Alzheimer (nonprofit foundation) and Research, University Hospital Würzburg. A. Mishra, R.M., T.J., S.N., D.C.P., M. Koido, Q.L.G., Genoscreen (biotech company); H.L.L.’s participation in this project was part of a competitive M. Shi, Y.H., M.K.G., I.C., K.K., Yi-Ching Liaw, F.C.V., K. Lin., B.S.W., V. Srinivasasainagendra, L.P., contract awarded to Data Tecnica International by the National Institutes of Health to support G.C., M.R.C., L.T., R.A., G.V.R., N.H., Y.H.J., J.Q.T., V.A., J.C., M.N., C.Y., E.Y., M.J.K., A.J.L., R.L., T.A., open science research; M.A.N.’s participation in this project was part of a competitive contract N.D.A., M.K.B., T.M.B., D.A.B., J.C.B., C.B., S.B., A.C., P.M.R., K. Cho, Z.C., J.W.C., P.L.d., R.d.C., awarded to Data Tecnica International by the National Institutes of Health to support open M.E., L.E.F., M.I.G., N.C.G., V.G., J. Hata, J. He., A.K.H., Y. Ho., A.S.H., H.I.H., M.I., M.A.J., C.E.J., C.J., science research, he also currently serves on the scientific advisory board for Clover M. Kamouchi, K.L.K., T. Kitazono, S.J.K., A.K., P.L., L.J.L., K. Lee, K. Lepik, J.L., L.L., A. Manichaikul, Therapeutics and is an advisor to Neuron23; N.A.M. declares institutional research grants to H.S.M., T. Meitinger, B.D.M., T. Morisaki, T.H.M., B.G.N., M.J.O., Y.O., N.C.O., B.O., A.P., S.S.R., J.R., the TIMI Study Group at Brigham and Women’s Hospital from Amgen, Pfizer, Ionis, Novartis, M.S.S., R.L.S., D. Saleheen, E.C.S., V. Salomaa, M. Sargurupremraj, M. Sasaki, C.L.S., C.O.S., A.S., AstraZeneca and NIH. The TIMI Study Group has received institutional research grant support N.L.S., K.S., Y.S., Y.V.S., K.T., S. Tiedt, T.T., N.P.T., H.K.T., D.T., S. Trompet, A.M.T., A.T., M.v.V., R.V., through Brigham and Women’s Hospital from Abbott, Amgen, Anthos Therapeutics, ARCA S.S.V., K.L.W., P.W., D.W., P.W.W., H.X., Q.Y., K.Y., I.Y.M., C.G., T.N., J.W.J., I.L.R., D. Strbian, Y.J.K., P.C., Biopharma, AstraZeneca, Daiichi-Sankyo, Eisai, Intarcia, Ionis Pharmaceuticals, MedImmune, E.M., M.R.I., H.A., S.W., K. Christensen, M.A.I., T.R., B.B.W., G.M.L., M.R., E.M.S., J.K., P.H.F., R.Z., Merck, Novartis, Pfizer, Regeneron Pharmaceuticals, Roche, The Medicines Company, Zora K.P., R.F., F.d.L., T.L., Y.M.R., W.T.L., K.J.J., L.B., G.P., D.I.C., J.I.R., J.Z., T.J.N., M.F., Yung-Po Liaw, Biosciences, Janssen Research and Development, Siemens Healthcare Diagnostics and I.F., R.G.W., M.O.O., J.E.H., L.M., Y.K., M.D. and S. Debette declare no competing interests. Softcell Medical; F.K.K. declares that the TIMI Study Group has received institutional research grant support through Brigham and Women’s Hospital from Abbott, Amgen, Anthos Therapeutics, ARCA Biopharma, AstraZeneca, Daiichi-Sankyo, Eisai, Intarcia, Ionis Additional information Pharmaceuticals, MedImmune, Merck, Novartis, Pfizer, Regeneron Pharmaceuticals, Roche, Supplementary information The online version contains supplementary material available at The Medicines Company and Zora Biosciences; M.S.S. has consultancies with Althera, Amgen, https://doi.org/10.1038/s41586-022-05165-3. Anthos Therapeutics, AstraZeneca, Beren Therapeutics, Bristol-Myers Squibb, DalCor, Correspondence and requests for materials should be addressed to Martin Dichgans or Dr. Reddy’s Laboratories, Fibrogen, IFM Therapeutics, Intarcia, Merck, Moderna, Novo Nordisk Stephanie Debette. and Silence Therapeutics, and research grant support through Brigham and Women’s Hospital Peer review information Nature thanks Paul Timmers and the other, anonymous, reviewer(s) from Abbott, Amgen, Anthos Therapeutics, AstraZeneca, Bayer, Daiichi-Sankyo, Eisai, Intarcia, for their contribution to the peer review of this work. Peer reviewer reports are available. Ionis, Medicines Company, MedImmune, Merck, Novartis, Pfizer, Quark Pharmaceuticals; C.T.R. Reprints and permissions information is available at http://www.nature.com/reprints. Article Extended Data Fig. 1 | GIGASTROKE study workflow. Study workflow and quantitative trait loci Mendelian Randomization; SuSiE: sum of single effects rationale. EUR: European; EAS: East-Asian; AFR: African; HIS: Hispanic; SAS: model; MENTR: Mutation Effect prediction on Non-coding RNA TRanscription; South Asian; AS: any stroke; AIS: any ischaemic stroke; LAS: large artery stroke; PIP: posterior probability; FDR: false discovery rate; LDSC-COV: covariate- CES: cardioembolic stroke; SVS: small vessel stroke; GWAS: genome-wide adjusted LD score regression; MR-Egger: Mendelian randomization-Egger; association study; IVW: inverse-variance weighted; MR-MEGA: meta-regression GREP: genome for REPositioning drugs; ATC: Anatomical Therapeutic Chemical; of multi-ethnic genetic association; COJO:conditional and joint analysis; P+T: pruning and thresholding; PRScs: polygenic risk score under continuous VEGAS2:versatile gene-based association study 2; MTAG: multi-trait analysis of shrinkage; BBJ: Biobank Japan; TIMI: thrombolysis in myocardial infarction; GWAS; TWAS: Transcriptome-wide association study; coloc: Colocalisation MVP: Million Veteran Program; SIREN: Stroke Investigative Research and Test; PWAS: Proteome-wide association studies; pQTL-MR: protein Educational Network. Extended Data Fig. 2 | Graphical representation of replication results. stroke and (B) cross-ancestry discovery and replication association results for Shown are effect sizes and 95% confidence intervals for the 48 replicated IVW any ischaemic stroke. OR = odds ratio; 95% CI = 95% confidence interval. loci for (A) cross-ancestry discovery and replication association results for any Article Extended Data Fig. 3 | Increase in power with increase in population Strat2_Strat1&EUR: N = 32,980 stroke cases, adding German and Dutch samples; diversity. The scatter plot shows the number of loci identified with Strat3_Strat2&EUR: N = 42,783, adding large European biobanks; Strat4_Strat3& incremental increase in sample size and population diversity. The diagonal EUR: N = 73,652 stroke cases, adding MEGASTROKE European; Strat5_Strat4& line reflects the increase in number of genome-wide significant loci with EAS: N = 91,303 stroke cases, adding Japanese BBJ; Strat6_Strat5&EAS: N = 99, increasing sample size of European ancestry only. The increase in number of 661 stroke cases, adding Chinese CKB; Strat7_Strat6&EAS: N = 101,065 stroke loci departs from this line when adding non-European ancestry samples; EUR: cases, adding other East Asian samples; Strat8_Strat7&AFR: N = 105,026 stroke European; EAS: East Asian; AFR: African; HIS: Hispanic; SAS: South Asian; cases, adding African ancestry samples; Strat9_Strat8&HIS: N = 106,542 BBJ: Biobank Japan; CKB: China Kadoorie Biobank. Strat0_EUR: N = 22,634 stroke cases, adding South American ancestry samples; Strat10_Strat9&SAS, stroke cases (European population-based longitudinal cohorts); Strat1_ N = 110,182 stroke cases, adding South Asian ancestry samples. Strat0&EUR: N = 26,253 stroke cases, adding Spanish and Estonian samples; Extended Data Fig. 4 | Association of stroke risk variants with vascular risk traits. We report only associations for which the stroke lead variant of a proxy in very high LD (r2 > 0.9) showed genome-wide significant association with the vascular risk trait in a prior GWAS. Colours represent the Z-scores of association of stroke risk increasing alleles with the trait. Article Extended Data Fig. 5 | Genetic correlations and Mendelian randomization significantly different from zero at a P < 0.05 shown as a full-sized square. causal estimates of 12 vascular risk factors and disease traits with stroke Genetic correlations or causal estimates that are significant after multiple (any and stroke subtypes), in European ancestry participants. Larger testing correction (P < 4.17 × 10 −3) are marked with an asterisk. Two-sided squares correspond to more significant P-values, with genetic correlations or P-values were calculated using LD score regression (LDSC) for genetic Mendelian randomization (MR) causal estimates (expressed in Z-scores) correlations and inverse variance weighted analysis for MR. Extended Data Fig. 6 | Genetic correlations and Mendelian randomization zero at a P < 0.05 shown as a full-sized square. Genetic correlations or causal causal estimates of 10 vascular risk factors and disease traits with stroke estimates (expressed in Z-scores) that are significant after multiple testing (any and stroke subtypes), in East Asian ancestry participants. Larger correction (P < 5 × 10 −3) are marked with an asterisk. Two-sided P-values were squares correspond to more significant P-values, with genetic correlations or calculated using LD score regression (LDSC) for genetic correlations and Mendelian randomization (MR) causal estimates significantly different from inverse variance weighted analysis for MR. CPD: cigarettes per day. Article Extended Data Fig. 7 | Transcriptome-wide association study of stroke in significant (p < 0.05) and COLOC PP4 ≥ 0.75; Genes are presented on the x-axis, multiple tissues. Heatmap of the transcriptome-wide association studies those underlined in blue are in a stroke GWAS locus, those underlined in purple (TWAS) of stroke (any stroke and stroke subtypes) showing transcriptome-wide are not within a genome-wide significant stroke risk locus (Methods); Tissue significant associations with supporting evidence from colocalization; types are on the y-axis (blue: cross-tissue weights; pink: arterial; orange: heart; Coloured squares are TWAS significant associations based on two-sided green: brain). p-values after multiple testing correction (p < 2.0 × 10 −6); * Conditionally (A) Single-cell gene expression data of TWAS-COLOC genes in dorsolateral (C) Cell-type enrichment in human and mouse single prefrontal cortex (ROS-MAP study) cell RNA-seq databases using STEAP Percent Expressed 9 VLMC 0 SMCs and pericytes 10 8 Cerebellum-Vascular endothelial cells 20 Cerebrum-Vascular endothelial cells OPC Intersection size 30 6 Identity oligodendrocytes Liver endothelial cell of Average Expression hepatic sinusoid VSMCA microglia 2 4 1 Astrocytes endothelial cells 0 2 2 2 astrocytes 2 -1 1 1 1 0 (B) Proteome-wide association study (PWAS) of stroke in brain tissue 2 CES Genes p value 5 AIS AS ICA1L 8.25e-10 ROS-MAP study 6 AS 4.21e-07 Banner study 12 LAS AIS ICA1L 1.01e-09 10 0 1.02e-06 0.0001 0.001 0.01 0.1 1 2 Odds ratio Extended Data Fig. 8 | Single-nucleus gene expression/enrichment analysis (C) Cell-type enrichment in human and mouse single cell RNA-seq databases and proteome-wide association study (PWAS) of stroke in brain tissue. using STEAP; the UpSet plot displays the number of significant enrichment (A) Single-nucleus gene expression data of TWAS-COLOC genes in dorsolateral results, by stroke subtype (horizontally; 2 for CES, 5 for AIS, 6 for AS, and 12 for prefrontal cortex (ROS-MAP study)21; Dot plot of the mean expression level in LAS) and by cell subtype (vertically; 2 cell-types show significant enrichment in expressing cells (colour) and percent of expressing cells (circle size) of selected LAS, AIS, and AS, 2 cell-types in AIS and AS, and 1 cell-type in LAS and AS, while genes across different cell types; (B) Proteome-wide association study (PWAS) 9 cell-types show significant enrichment in LAS only, 2 in CES only and 1 in of stroke in brain tissue; Box plot showing effect estimates (odds ratio) for AS and AIS respectively); details are displayed in Supplementary Table 29. AS: associations of pQTL of ICA1L in the ROS-MAP (N = 376 independent samples) any stroke; AIS: any ischaemic stroke; LAS: large artery stroke; CES: cardioembolic and Banner (N = 152 independent samples) studies with any stroke (AS) and any stroke; VLMC: vascular and leptomeningeal cells, OPC: oligodendrocyte ischaemic stroke (AIS), identified in PWAS after multiple testing correction. progenitor cells, SMC: smooth muscle cells; VSMCA: vascular smooth muscle Odds ratios ± 95% CIs are shown. Dashed line indicates an odds ratio of 1. cells, arterial. Two-sided p-values were computed using the TWAS-COLOC approach. Article Extended Data Fig. 9 | Drug target pQTL PheWAS. PheWAS in Estonian Two-sided P-values were calculated using logistic regression to test association biobank for rs2289252, a cis-pQTL of F11. Each triangle in the plot represents between the pQTL and Phecodes (p = 3.45 × 10 −6 for phenome-wide one Phecode and the direction of the triangle represents direction of effect. significance). (A) Derivation of standard PGS models Europeans East Asians LD reference panel LD reference panel GWAS summary statistics GWAS summary statistics (1000 Genomes EUR; n=503) (1000 Genomes EAS; n=503) Compute candidate PGS models using P+ T, LDpred, and PRS-CS Compute candidate PGS models using P+T, LDpred, and PRS-CS Candidate PGS models Candidate PGS models (37 models) (37 models) Model selection based on maximal AUC Model selection based on maximal AUC (EUR case-control data with 1,003 AIS cases and 8,997 controls) (EAS case-control data with 577 AIS cases and 9,232 controls) Selected best PGS model Selected best PGS model (1 model) (1 model) (B) Derivation and evaluation of integrative PGS models Stroke outcomes (European-specific GWAS) Stroke outcomes (cross-ancestry GWAS) Strake outcomes (East Aisan-specific GWAS) Stroke outcomes (cross-ancestry GWAS) AS AIS LAS SVS CES AS AIS LAS SVS CES AS AIS LAS SVS CES AS AIS LAS SVS CES Stroke-related phenotypes (European-specific GWAS) Stroke-related phenotypes (East Asian-specific GWAS) AF CAD T2D SBD DBD TC LDL-C HDL-C TG BMI Height Smoking AF CAD T2D SBD DBD TC LDL-C HDL-C TG BMI Height Smoking Model training using elastic-net logistic regression Pre-filtering and model training using elastic-net logistic regression (EUR case-control data with 1,003 AIS cases and 8,997 controls) (EAS case-control data with 577 AIS cases and 9,232 controls) Integrative PGS model for EUR Integrative PGS model for EAS (1 model) (1 model) Evaluation of the integrative PGS model Evaluation of the integrative PGS model (EUR prospective data with 102,099 subjects including 1,128 incident AIS cases) (EAS case-control data with 1,470 AIS cases and 40,459 controls) Extended Data Fig. 10 | Derivation and evaluation of integrative polygenic an East-Asian iPGS model consisting of 6,010,730 variants were constructed by score models for Europeans and East Asians. (A) With summary statistics combining the 11 and 7 significant PGS models using the elastic-net derived of 22 GWAS (10 GIGASTROKE and 12 on vascular risk factors) and linkage weights respectively. The European and East Asian iPGS models were evaluated disequilibrium reference data of 1000 Genomes Europeans (n = 503) and in the European (a European prospective cohort data with 102,099 participants East Asians (n = 504), we computed 37 candidate PGS models using P+T, including 1,128 incident IS cases) and East-Asian (AIS case-control data, Ncases/ LDpred, and PRScs algorithms. For each GWAS, the best PGS model was Ncontrol = 1,470/40,459) model evaluation dataset (Methods); AS indicates any selected based on the maximal area under the curve (AUC) values in the stroke; AIS, any ischaemic stroke; LAS, large artery stroke; SVS, small vessel training dataset of Europeans (any ischaemic stroke [AIS] case-control data, stroke; CES, cardioembolic stroke; AF, atrial fibrillation; CAD, coronary artery Ncases/Ncontrols = 1,003/8,997) and East Asians (AIS case-control data, disease; T2D, type 2 diabetes; SBP, systolic blood pressure; DBP, diastolic blood Ncases/Ncontrols = 577/9,232). Out of 22 selected PGS models derived from the pressure; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; 22 GWAS, 11 and 7 were significantly associated with AIS in the European and HDL-C, high-density lipoprotein cholesterol; TG, triglyceride; BMI, body mass East Asian training dataset respectively (Bonferroni-corrected P < 0.05). index; AUC indicates area under the curve; EUR, European; EAS, East Asian; (B) The significant PGS models were used as the variables for elastic-net logistic GWAS, genome-wide association study; LD, linkage disequilibrium; PGS, regression and the weights for the variables were trained using the model polygenic score. training dataset. The European iPGS model consisting of 1,213,574 variants and