Bradbury et al. Journal of Translational Medicine (2022) 20:611 Journal of https://doi.org/10.1186/s12967-022-03816-7 Translational Medicine RESEARCH Open Access A combination of molecular and clinical parameters provides a new strategy for high‑grade serous ovarian cancer patient management Melissa Bradbury1,2,3,4, Eva Borràs1,2, Marta Vilar1,3, Josep Castellví5, José Luis Sánchez‑Iglesias3,4, Assumpció Pérez‑Benavente3,4, Antonio Gil‑Moreno3,4,6, Anna Santamaria3,7*† and Eduard Sabidó1,2*†    Abstract Background: High-grade serous carcinoma (HGSC) is the most common and deadly subtype of ovarian cancer. Although most patients will initially respond to first-line treatment with a combination of surgery and platinum-based chemotherapy, up to a quarter will be resistant to treatment. We aimed to identify a new strategy to improve HGSC patient management at the time of cancer diagnosis (HGSC-1LTR). Methods: A total of 109 ready-available formalin-fixed paraffin-embedded HGSC tissues obtained at the time of HGSC diagnosis were selected for proteomic analysis. Clinical data, treatment approach and outcomes were collected for all patients. An initial discovery cohort (n = 21) were divided into chemoresistant and chemosensitive groups and evaluated using discovery mass-spectrometry (MS)-based proteomics. Proteins showing differential abundance between groups were verified in a verification cohort (n = 88) using targeted MS-based proteomics. A logistic regres‑ sion model was used to select those proteins able to correctly classify patients into chemoresistant and chemosensi‑ tive. The classification performance of the protein and clinical data combinations were assessed through the genera‑ tion of receiver operating characteristic (ROC) curves. Results: Using the HGSC-1LTR strategy we have identified a molecular signature (TKT, LAMC1 and FUCO) that combined with ready available clinical data (patients’ age, menopausal status, serum CA125 levels, and treatment approach) is able to predict patient response to first-line treatment with an AUC: 0.82 (95% CI 0.72–0.92). Conclusions: We have established a new strategy that combines molecular and clinical parameters to predict the response to first-line treatment in HGSC patients (HGSC-1LTR). This strategy can allow the identification of chemore‑ sistance at the time of diagnosis providing the optimization of therapeutic decision making and the evaluation of † Anna Santamaria and Eduard Sabidó authors contributed equally *Correspondence: anna.santamaria@vhir.org; eduard.sabido@crg.cat 1 Centre de Regulació Genòmica, Barcelona Institute of Science and Technology (BIST), Dr Aiguader 88, 08003 Barcelona, Spain 3 Biomedical Research Group in Gynecology, Vall d’Hebron Institut de Recerca, Universitat Autònoma de Barcelona, Vall, d’Hebron Barcelona Hospital Campus, Passeig Vall d’Hebron 119‑129, 08035 Barcelona, Spain Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​ mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 2 of 13 alternative treatment strategies. Thus, advancing towards the improvement of patient outcome and the individualiza‑ tion of HGSC patients’ care. Keywords: High-grade serous ovarian cancer, Proteomics, Biomarker, Prediction, Treatment Background alternative treatment strategies to be considered at the High-grade serous carcinoma (HGSC) remains the most time of cancer diagnosis (i.e. non-platinum drugs) thus common and deadly subtype of ovarian cancer, due to improving patients’ response and survival. It also opti- its diagnosis at advanced stages in over 80% of cases [1]. mizes therapeutic decision making and allows an indi- The standard of care of advanced stage HGSC is a com- vidualized management strategy by avoiding the use of bination of cytoreductive surgery and platinum-based futile treatments, thus improving patients’ quality of life. chemotherapy (i.e. carboplatin and paclitaxel). Cytore- This strategy is therefore of special relevance for patients ductive surgery can be performed at the time of cancer with advanced HGSC for whom a predictive method is diagnosis or can be delayed after 3 or 4 cycles of neoad- lacking. Thus, in this work we aim to identify a protein juvant chemotherapy [2]. Recently, the use of a targeted biomarker signature able to predict response to first-line management approach (i.e. anti-angiogenic agents and treatment in patients with newly diagnosed HGSC. PARP inhibitors) has been introduced for specific sub- groups of patients such as those with BRCA mutations Methods [3, 4]. Although the results on BRCA mutation status can Patient cohorts be available within 2–3 weeks from diagnosis in some A total of 109 patients with newly diagnosed advanced referral oncological centres, in most clinical settings stage HGSC managed at the Gynecological Oncology worldwide it is not available until first-line treatment Unit at the Hospital Vall d’Hebron (Barcelona, Spain) has already commenced. Although HGSC is considered between 1996 and 2017 were included in the study. Out to be a platinum-sensitive disease, ~ 20–30% of patients of these, a fifth (21 patients) were randomly selected fail to respond or experience disease recurrence within for the discovery cohort and four fifths (88 patients) for 6 months of completing chemotherapy [5]. In clinical the verification cohort. Patients were classified accord- practice, these patients are considered to be resistant ing to their treatment-free interval to platinum (TFIp) to platinum, and second-line treatment is usually based into two main groups: TFIp < 6 months (chemoresistant) on non-platinum agents (e.g. gemcitabine, pegylated and TFIp > 6 months (chemosensitive). For the discovery liposomal doxorubicin, topotecan or weekly paclitaxel) cohort, the chemosensitive group was subdivided into and control of symptoms [2, 6]. Efforts to establish bio- those patients who developed recurrence (chemosensi- logical stratification profiles in HGSC are mainly being tive with TFIp > 6 months) and those who did not recur focused on genomic and transcriptomic markers [7–10], (chemosensitive with no recurrence) in order to cat- while the evaluation of protein markers to assess treat- egorise the subgroup of patients with better response to ment response and/or survival in HGSC tissue samples chemotherapy. The characteristics of all patients included remain scarce [11–13]. Beyond their clinical relevance in the discovery and verification phases are summarized as biological endpoints and drug targets, proteins are in Tables 1 and 2 and Additional file 1: Table S1 and markers widely used in clinical tests for disease diagnosis Additional file 2: Table S2. and prognosis [13, 14]. In addition, the use of proteomic approaches (i.e. mass spectrometry) enable the system- Formalin‑fixed paraffin‑embedded tissue samples atic interrogation of proteomes from complex clinical Formalin-fixed paraffin-embedded (FFPE) tissue samples samples which can later be translated into immunoassays were obtained from the Pathology Department human for clinical use [15, 16]. Despite the numerous studies tissue repository at the Hospital Vall d’Hebron (Barce- being carried out to identify novel protein biomarkers in lona, Spain). All cases corresponded to HGSC tumour ovarian cancer, there are currently no validated markers biopsies obtained at the time of diagnosis. Samples were used in the clinic to predict response to first-line treat- registered, processed and fixed following the hospital ment and guide the management of patients with newly standard operating procedures. Pathological and clini- diagnosed HGSC. In this work, we present a new strategy cal data from all gynaecological registered samples were that combines molecular and routine clinical parameters manually reviewed to ensure that all patients included to anticipate the response to first-line treatment in HGSC in the study met the following criteria: (a) newly diag- patients (HGSC-1LTR). Prediction of those patients nosed advanced stage HGSC (stages III and IV), (b) treat- less likely to respond to first-line chemotherapy allows ment with primary surgery and six cycles of carboplatin Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 3 of 13 Table 1 Summary of the clinical characteristics of the patient groups included in the discovery cohort Chemoresistant with TFIp < 6 m Chemosensitive with TFIp > 6 m Chemosensitive with (n = 7) (n = 7) no recurrence (n = 7) Age (years) Average (range) 67 (53–77) 58 (49–69) 60 (43–70) Menopausal status (n, %) Premenopausal 0 2 (29%) 1 (14%) Postmenopausal 7 (100%) 5 (71%) 6 (86%) Stage (n, %) IIIC 6 (86%) 6 (86%) 7 (100%) IV 1 (14%) 1 (14%) 0 Serum CA125 levels at diagnosis (U/mL) Average (range) 1944 (809–5500) 2444 (126–6992) 1051 (100–2856) TFIp (months) Average (range) 3 (1–5) 28 (11–67) NA Primary treatment (n, %) Cytoreductive surgery 4 (57%) 5 (71%) 4 (57%) Neoadjuvant chemotherapy 3 (43%) 2 (29%) 3 (43%) Status (n, %) Alive 0 1 (14%) 7 (100%) Dead 7 (100%) 6 (86%) 0 TFIp: treatment-free interval to platinum Table 2 Summary of the clinical characteristics of the patient groups included in the verification cohort Chemoresistant (n = 25) Chemosensitive (n = 63) Age (years) Average (range) 63 (49–81) 59 (37–84) Menopausal status (n, %) Premenopausal 1 (4%) 18 (29%) Postmenopausal 24 (96%) 45 (71%) Stage (n, %) IIIA 0 2 (3%) IIIB 1 (4%) 6 (10%) IIIC 16 (64%) 45 (71%) IVA 3 (12%) 4 (6%) IVB 5 (20%) 6 (10%) Serum CA125 levels at diagnosis (U/mL) Average (range) 2246 (161–10,548) 1385 (20–19,007) TFIp (months) Average (range) 3 (1–5) 13 (6–63) Primary treatment (n, %) Cytoreductive surgery 7 (28%) 44 (70%) Neoadjuvant chemotherapy 18 (72%) 19 (30%) Status (n, %) Alive 0 31 (49%) Dead 25 (100%) 32 (51%) TFIp: treatment-free interval to platinum Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 4 of 13 and paclitaxel chemotherapy, (c) follow-up until disease selected according to at least one of the following crite- recurrence or at least five years after treatment comple- ria: (a) were present in at least 4 patients in one group tion. The histopathological features of the selected sam- and completely absent in the comparison group, (b) were ples were then reviewed by an experienced pathologist either completely present or completely absent in the to confirm the diagnosis and tumour content. Areas chemoresistant group, (c) had been previously reported containing 80% or more tumour with no areas of necro- in the literature as potential predictive markers in ovar- sis were selected for serial sectioning and further sample ian cancer. Nine proteins were included using these crite- processing. ria, three of which corresponded to proteins drawn from the literature (i.e. CT45, CDK1 and CLDN3) [11, 21–23]. Discovery cohort Sample preparation for proteomic analysis Verification cohort An initial cohort of 21 patients were randomly selected Sample preparation for proteomic analysis for biomarker discovery from whom FFPE tumour sam- A cohort of 88 patients were randomly selected for bio- ples were collected. For FFPE sample preparation we marker verification from whom FFPE samples were adapted a recently described workflow capable of yield- collected. FFPE tissues were cut in 5 serial sections of ing substantial amounts of peptides for quantification by 10 µM thick using a microtome. A clean blade was used proteomic analysis with high reproducibility [17]. FFPE for sectioning each tissue sample separately. Tissue sec- tissues were cut in five serial sections of 10 µm thick tions were deparaffinated in 1 ml xylene (3 min at 50 °C) using a microtome. A clean blade was used for section- and washed twice with 1 ml absolute ethanol. Ethanol ing each tissue sample separately. Tissue sections were was removed completely and sections were left to air- deparaffinated in 1 ml xylene (3 min at 50 °C) and washed dry. Samples were resuspended in lysis buffer (40 mM twice with 1 ml absolute ethanol. Ethanol was removed TrisHCl, 1% SDS, pH 8.2) and left to incubate at 99 °C completely and sections were left to air-dry. for 30 min and 80 °C for 2 h in a thermomixer. Following centrifugation (20 min, RT, 15,000g) the supernatant was Data analysis quantified using BCA protein assay. Protein extracts were Acquired data were analysed using the Proteome Discov- diluted 20 times with 50 mM ammonium bicarbonate for erer software suite (v.2.0, Thermo Fisher Scientific) and digestion with trypsin (1:50 w:w, 37 °C, 8 h, Promega cat using the HIPPR™ Detergent Removal Spin Column Kit peptides were identified using the Mascot search engine # V5113). Detergent was removed from protein digests (v.2.5.1, Matrix Science). Data were searched against the Swiss-Prot human protein database (as in October (Thermo Scientific, PN 88305) following manufacturer 2017, 20,239 entries) plus a list of common contaminants instructions. Peptide mix was acidified with formic acid (148 entries) [18]. The precursor ion mass tolerance was 5% and desalted with a MicroSpin C18 column (The Nest 7 ppm at the MS1 level, and up to three missed cleavages Group, Inc). Isotopically-labelled peptides (13C6,15N2-Lys for trypsin were allowed. The fragment ion mass toler- and 13C6,15N4-Arg, Pepotec Peptides, Thermo Fisher Sci- ance was set to 0.5 Da and methionine oxidation was set entific) were spiked in the peptide mixtures and used as as variable modification. The identified peptides were fil- internal standard for quantification by parallel reaction tered by 5% FDR. Peptide areas were obtained using the monitoring (PRM). A total of 30 proteins and 59 peptides “Precursor Ions Area Detector” module in the Proteome were selected for PRM from 88 samples (25 chemoresist- Discoverer software suite (v.2.0, Thermo Fisher Scien- ant and 63 chemosensitive). tific). Protein abundance in each condition was estimated using the average of the three most intense peptides per Parallel reaction monitoring protein group (Additional file 3: Table S3). Data were Up to two unique peptides per protein were selected for log2-transformed, normalized by equalised median and targeted protein quantification, prioritizing those pep- quantified using the MSstats R software package v.3.8.2 tides that had been previously observed in the discovery [19]. Only those proteins with quantitative values in at cohort. For each selected peptide, an isotopically-labelled least 4 out of 7 patients per group were considered in the peptide (13C6,15N4-Arginine, and 13C6,15N2-Lys) was group comparison analysis. Changes in protein abun- spiked in the samples and used as an internal standard dance between groups were compared using two-sided t for quantification by Parallel reaction monitoring (PRM). test analysis followed by correction for multiple testing The amount of internal standard peptide to be spiked in [20]. Changes were considered significant with a q-value each sample was evaluated using dilution curves and the below 0.05. final concentration was chosen based on the following All significant differentially expressed proteins were criteria: (a) to be within the concentration range in which included in the analysis. Additional proteins were a linear response of the peptide was observed and, (b to Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 5 of 13 have an area as close to the endogenous peptide area as log2-transformed abundance for a given protein across possible. One microgram of digested sample was analysed runs. by PRM using an Orbitrap Fusion Lumos (Thermo Fisher Scientific) coupled to an EASY-nanoLC 1000 UPLC sys- Predictive analysis tem (Thermo Fisher Scientific) with a 50-cm C18 chro- For predictive analysis 29 proteins and 29 peptides (best matographic column. Peptide mixes were separated with peptide for each protein) were used (Additional file 6: a chromatographic gradient starting at 95% buffer A and Table S5). The final verification cohort (n = 80) was 5% buffer B with a flow rate of 300 nL/min and going up divided into a training subset and a validation subset to 25% buffer B and 75% A in 52 min and to 40% B and with a 8:10 ratio. Within the training set, the abundance 60% A in 8 min (Buffer A: 0.1% formic acid in water and of each protein was fitted in a logistic regression model Buffer B: 0.1% formic acid in acetonitrile). The Orbitrap between chemoresistant and chemosensitive patients and Fusion Lumos was operated in positive ionization mode the classification ability of each protein was evaluated with an EASY-Spray nanosource at 2.4 kV and at a source by the area under the curve (AUC) of a receiver operat- temperature of 275 °C. A scheduled PRM method was ing characteristic. The protein with the highest AUC was used for data acquisition with a quadrupole isolation win- selected as the first classifier. Most discriminative pro- dow set to 1.4 m/z and MS2 scans over a mass range of teins were repeatedly added to the classifier one by one m/z 340–950, with detection in the Orbitrap mass ana- as long as their combination resulted in an increase in lyser at a 60 K resolution. MS2 fragmentation was per- AUC value higher than 0.02. The best protein combina- formed using HCD fragmentation at normalised collision tion in the training subset was fitted in a logistic regres- energy of 30%, the AGC was set at 50,000 and the maxi- sion model and was applied to the validation subset. The mum injection time at 118 ms. The size of the sched- procedure from division into training and validation set uled window was 6 min and the maximum cycle time to fitting of the logistic model with the best classification was 2.8 s. All data was acquired with XCalibur software signature was repeated 500 times to assess the reproduc- v.3.0.63. QCloud was used to control instrument longitu- ibility of classification ability. The final consensus model dinal performance during the project [24]. was comprised of the combination of proteins which were selected with higher frequencies in the 500 repeats Data analysis [26]. The pROC package in R was used to draw ROCs, Product ion chromatographic traces corresponding to the calculate AUCs and other predictive performance data targeted precursor peptides were evaluated with Skyline including sensitivity and specificity at the optimal cut-off software v.4.2 [25] based on: (a) the number of detected threshold (Youden J Index) for discrimination between traces, (b) co-elution of endogenous traces, (c) co-elution groups [27]. of endogenous and internal standard peptides, (d) cor- relation of the trace relative intensities between endog- Results enous and internal standard peptides and, (e) expected Identification of protein biomarker candidates retention time. Those transitions showing interferences from formalin‑fixed paraffin embedded HGSC tissues on the PRM traces were discarded. Measurements which Our first aim was to identify potential protein biomarker were considered to be under the limit of detection were candidates able to predict response to first-line treat- replaced with an estimation of the background value. ment in patients with newly diagnosed HGSC. To this Peptides who did not fulfil all the above criteria were intent, we performed a discovery proteomic analysis of removed from the study (seven peptides in total). In FFPE tumour samples from a cohort of 21 patients diag- addition, eight samples were also discarded due to sam- nosed with advanced stage HGSC. This analysis identi- ple quality. In all, we consistently quantified a total of 52 fied proteins that differed in abundance in relation to the peptides corresponding to 29 proteins across 80 samples. patients’ response to chemotherapy with carboplatin and Peak areas were obtained for each production and data paclitaxel. We divided the cohort according to the time of were log2-transformed prior to normalization and statis- disease recurrence into chemoresistant and chemosensi- tical analysis (Additional file 5: Table S4). Normalization tive [28] (see “Materials and Methods” section). Table 1 relied on internal isotopically-labelled standard peptides and Additional file 1: Table S1 summarize the clinical which were used to equalise the median abundance of characteristics of the groups of patients included in the the internal standard peptides across all runs and then discovery cohort. shift all endogenous areas in a run by a same amount. We quantified a total of 6813 proteins in our discov- Protein abundance estimates were performed with the ery dataset (Fig. 1B). Only those proteins identified in software package MSstats 3.14.1 [19]. Missing quantifi- at least 4 out of 7 patients in each group (2441 proteins cation values were imputed with a minimum estimated on average) were considered for quantitative analysis Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 6 of 13 Fig. 1 Identification of protein biomarker candidates using discovery proteomics. A Graphical abstract of the patient groups and number of samples included in the discovery phase. B Overview of the proteomics workflow used in the discovery phase. HGSC indicates high-grade serous ovarian cancer tumours. C Volcano plots of the pairwise comparisons between the three groups included in the discovery phase (chemoresistant with TFIp < 6 months, chemosensitive with TFIp > 6 months and chemosensitive with no recurrence). Highlighted in red are the differentially expressed proteins and highlighted in blue are the proteins present in at least four patients in one group and completely absent in the other group and statistical group comparison evaluation (Additional 30 proteins were selected as potential protein biomarker file 4: Fig. S1A). We identified 27 differentially expressed candidates for verification by targeted proteomics in an proteins between the three groups (chemoresistant with independent cohort of HGSC patients. TFIp < 6 months, chemosensitive with TFIp > 6 months, Verification of protein biomarker candidates and chemosensitive with no recurrence) (q-value < 0.05) from formalin‑fixed paraffin‑ embedded HGSC tissues (Fig. 1C). Of these, 21 proteins showed changes in abun- Protein verification acts as a bridging phase capable dance, whilst 6 proteins were present in at least four of overcoming the gap between biomarker discovery patients in one group and completely absent in the com- and validation [15, 29]. In the verification cohort we parison group. These six proteins include CISD2, CRP, aimed to accurately quantify the candidate biomark- DNAJC10, ID4, SYUA and protein C8orf33. Additionally, ers obtained by discovery proteomics as a basis for our three proteins previously shown to be associated to ovar- predictive analysis circumscribing the intended use ian cancer prognosis were added from the literature (i.e. of our potential predictive biomarker signature to the CT45, CDK1 and CLDN3) [11, 21–23]. Hence, a total of Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 7 of 13 classification of patients into the chemoresistant and Identification of a protein signature able to classify chemosensitive groups. patients into chemoresistant and chemosensitive groups We verified the biomarker candidates selected from The final step of this verification cohort was aimed at the discovery cohort using targeted proteomics. In par- defining protein combinations able to classify patients ticular, we used parallel reaction monitoring (PRM) in diagnosed with advanced stage HGSC into those who an independent cohort of 88 FFPE tumour samples to will respond to first-line chemotherapy treatment with assess the ability of the selected biomarker candidates carboplatin and paclitaxel and those who will not. To this to predict patients’ response to first-line treatment. In intent, we performed a logistic regression model to select this verification cohort, we also sought to define the and evaluate those proteins able to correctly classify potential clinical applicability of a protein biomarker patients into chemoresistant and chemosensitive groups, signature. Because the intended use is the prediction of both individually and in combination. Details of the response to first-line treatment in patients with newly cross-validation analysis can be found in the correspond- diagnosed advanced stage HGSC, we limited our pre- ing methods section. In brief, the final patient cohort dictive analysis to two groups based on the TFIp cut- (n = 80) was randomly divided into a training set and off of 6 months and therefore divided our verification validation set. The classification power of each protein cohort into chemoresistant (TFIp < 6 months) (n = 25) was first evaluated in the training set by a logistic regres- and chemosensitive (TFIp > 6 months) (n = 63) groups sion model and additional proteins were then added into (Fig. 2A). The characteristics of the patient groups the best protein classifier in a stepwise manner. The vali- included in the verification phase are summarized in dation set was then used to evaluate the discriminatory Table 2 and Additional file 2: Table S2. performance between chemoresistant and chemosensi- FFPE tumour samples were processed in the same tive patients. This cross-validation process was repeated manner as in the discovery phase. In order to quantify 500 times in order to assess its robustness. The classifi- the protein biomarker candidates by PRM, we selected cation performance of the most frequently selected pro- a specific subset of representative peptides which tein combinations were assessed within the whole dataset were used as surrogates for each protein candidate. through the generation of receiver operating character- We chose 1 or 2 peptides per protein based on their istic (ROC) curves. This predictive analysis identified a uniqueness to the given target, their chromatographic 3-protein combination including transketolase (TKT, and mass spectrometry performance and their stabil- P29401), laminin subunit gamma-1 (LAMC1, P11047) ity [30]. Using these criteria we selected a total of 59 and tissue alpha-L-fucosidase (FUCO, P04066), as the peptides corresponding to the 30 protein candidates best protein classifier of chemotherapy response with for quantification by PRM (Table 3). Internal standard an AUC of 0.76 (95% CI 0.64–0.87). This protein combi- peptides were used to guide the identification of the nation was followed by another 3-protein combination endogenous peptides and to accurately quantify them which contained TKT, LAMC1 and c-reactive protein in the tumour samples. These standards consisted of (CRP, P02741) with an AUC of 0.75 (95% CI 0.64–0.86) identical peptide sequences as the endogenous with (Fig. 2B, C Additional file 7: Table S6 and Additional the C-terminus amino acid isotopically labelled which file 8: Fig. S2A, B). were spiked into each sample for analysis (Additional Because response rates of platinum sensitivity are file 4: Fig. S1B). The technical variability of the assay known to fall in a continuum, we next assessed if was calculated as a coefficient of variation (CV) show- both protein combinations (TKT + LAMC1 + FUCO ing a median CV of 3%, being the highest value 26% and TKT + LAMC1 + CRP) were also good classi- (Additional file 4: Fig. S1C). fiers when considering the most marginal group of (See figure on next page.) Fig. 2 Predicted proteins able to discriminate between chemoresistant and chemosensitive patients. A Graphical abstract of the patient groups included in the verification cohort. B Frequency plot representing the number of times a protein combination was selected over 500 iterations. C Targeted proteomic profiles (fragment extracted ion chromatograms) corresponding to the endogenous peptides and their internal standards for proteins TKT, LAMC1 and FUCO. D Receiver operating curves corresponding to the predictor formed by the combination of proteins TKT + LAMC1 + FUCO and clinical parameters including age, menopausal status, CA125 levels at diagnosis and decision to treat with either primary cytoreductive surgery or neoadjuvant chemotherapy. The protein combination TKT + LAMC1 + FUCO (AUC from 0.76; 95% CI 0.64–0.87) in combination with clinical data (AUC 0.75; 95% CI 0.63–0.88) increased the AUC value to 0.82 (95% CI 0.72–0.92), p = 0.09. E Workflow of the HGSC-1LTR strategy. Identification of chemoresistance can facilitate the study of alternative treatments to improve patient outcome. Patients classified as chemosensitive could undergo the standard of care with platinum-based agents Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 8 of 13 Fig. 2 (See legend on previous page.) Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 9 of 13 Table 3 List of the 30 proteins and 52 peptides selected for Parallel Reaction Monitoring measurement Accession Protein Peptide Q8N6H7 ARFG2 _ADP-ribosylation factor GTPase-activating protein 2 EVDAEYEAR LAYQELQIDR Q8IXM2 BAP18 _Chromatin complexes subunit BAP18 VYEDSGIPLPAESPK VGEIFSAAGAAFTK Q9H7E9 C8orf33 _UPF0488 protein C8orf33 AAAYSAQVQPVDGATR​ LPGPVSSAR P06493 CDK1 _Cyclin-dependent kinase 1 SPEVLLGSAR NLDENGLDLLSK Q8N5K1 CISD2 _CDGSH iron-sulfur domain-containing protein 2 LPVPESITGFAR DSLINLK O15551 CLD3 _Claudin-3 STGPGASLGTGYDR VVYSAPR Q9UNS2 CSN3 _COP9 signalosome complex subunit 3 ASALEQFVNSVR AMDQEITVNPQFVQK P02741 CRP _C-reactive protein ESDTSYVSLK QDNEILIFWSK P0DMU7 CT45A6 _Cancer/testis antigen family 45 member A6 VAVDPETVFK IFEMLEGVQGPTAVR Q8IXB1 DJC10 _DnaJ homolog subfamily C member 10 NFQEEQINTR ILYDILAFAK P58107 EPIPL _Epiplakin AEAEAGSPRPDPR ALQQGLVGLELK P04066 FUCO _Tissue alpha-L-fucidase DLVGELGTALR FFHPEEWADLFQAAGAK P07093 GDN _Glia-derived nexin ASAATTAILIAR IEVSEDGTK P47928 ID4 _DNA-binding protein inhibitor ID-4 TPLTALNTDPAGAVNK P19823 ITIH2 _Inter-alpha-trypsin inhibitor heavy chain H2 LSNENHGIAQR IQPSGGTNINEALLR P11047 LAMC1 _Laminin subunit gamma-1 NTIEETGNLAEQAR EAQQALGSAAADATEAK O14910 LIN7A _Protein lin-7 homolog A IIPGGVAER ATVAAFAASEGHSHPR P62312 LSM6 _U6 snRNA-associated Sm-like protein LSm6 LNSGVDYR GNNVLYISTQK Q99733 NP1L4 _Nucleome assembly protein 1-like 4 AAATAEEPDPK VLAALQER P12955 PEPD_Xaa-Pro dipeptidase IDEPGLR LPASHATWMGK P28062 PSB8 _Proteasome subunit beta type-8 FQHGVIAAVDSR ASAGSYISALR O00584 RNT2 _Ribonuclease T2 VYGVIPK ELDLNSVLLK P23526 SAHH_Adenylhomocysteinase GISEETTTGVHNLYK VADIGLAAWGR​ O75533 SF3B1 _Splicing factor 3B subunit 1 THEDIEAQIR WDQTADQTPGATPK P37840 SYUA _Alpha-synuclein EGVVHGVATVAEK TVEGAGSIAAATGFVK Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 10 of 13 Table 3 (continued) Accession Protein Peptide P35269 T2FA _General transcription factor IIF subunit 1 LDTGPQSLSGK EFRPEDQPWLLR Q8WUY1 THEM6 _Protein THEM6 LLEPFEVR AFYLEAR P29401 TKT _Transketolase SVPTSTVFYPSDGVATEK ILATPPQEDAPSVDIANIR P04350 TBB4A _Tubulin beta-4A chain EVDEQMLSVQSK INVYYNEATGGNYVPR P62253 UB2G1 _Ubiquitin-conjugating enzyme E2 G1 FITEIWHPNVDK TELQSALLLR chemosensitive patients. Thus, we compared the abil- an AUC value of 0.82 (95% CI 0.72–0.92), p-value = 0.09 ity of our protein combinations to discriminate between (Fig. 2D). Interestingly, this analysis shows how the addi- chemoresistant and partially chemosensitive patients tion of molecular information, and more concretely the (i.e. those patients who develop recurrence between 6 protein signature identified in this work, can substan- and 12 months after the last dose of chemotherapy). tially improve its classificatory performance provided by These partially chemosensitive patients are still taken clinical data alone (Additional file 9: Table S7). into consideration in clinical trials assessing chemo- therapy response although recent evidence shows they Discussion also benefit from re-treatment with platinum-based In this work we have established a new strategy that regimens [31, 32]. We observed that protein combina- combines molecular and clinical parameters to predict tion TKT + LAMC1 + FUCO was able to discriminate the response to first-line treatment in HGSC patients between chemoresistant and partially chemosensitive (HGSC-1LTR) (Fig. 2E). There are currently no protein patients with an AUC of 0.76 (95% CI 0.61–0.91). Con- biomarkers available at the time of HGSC diagnosis able trary, TKT + LAMC1 + CRP had worse discriminatory to predict patients individual response to first-line chem- ability with an AUC value of 0.70 (95% CI 0.53–0.87) otherapy with carboplatin and paclitaxel. In addition, (Additional file 8: Fig. S2C). Therefore, the 3-protein studies evaluating predictive protein markers in ovarian combination signature TKT + LAMC1 + FUCO showed cancer tissues using proteomic approaches are scarce the best ability to classify patients into chemoresistant as highlighted in a recent review by our group [13]. and chemosensitive groups, even when considering the The identification of these predictors of chemotherapy closest subgroups of patients. response, as the one presented here, allows the prioritiza- tion of platinum-based agents if the disease is sensitive, The addition of clinical parameters improves and the use of alternative treatments if resistant in order the classificatory ability of the protein signature to improve patient management. This ability to predict Next, we assessed if the addition of clinical information a patient’s response is particularly relevant in chemore- could improve the classificatory performance of our pro- sistant HGSC because alternative non-platinum based tein biomarker signature. We included relevant clinical chemotherapy regimens could be considered or clinically data routinely used by clinicians during the assessment of evaluated in this subgroup of patients [33]. Addition- patients diagnosed with HGSC prior to treatment. This ally, it offers several clinical advantages and aid informed data included patients’ age, menopausal status, serum clinical decisions. Clinicians could address patients’ indi- CA125 levels at the time of diagnosis and the decision to vidual needs and improve their quality of life by avoid- treat with either primary cytoreductive surgery or neoad- ing the use of futile treatments. It could also alleviate the juvant chemotherapy following clinical, radiological and economic burden of the healthcare system associated surgical evaluation. The combined analysis of our pro- with the use of ineffective treatments. Although under- tein biomarker signature (AUC 0.76; 95% CI 0.64–0.87) standing why platinum-resistance occurs is essential for with clinical parameters (AUC 0.75; 95% CI 0.63–0.88) improving survival, new strategies able to discriminate showed that it provided a better classificatory power with between chemoresistant and chemosensitive patients at Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 11 of 13 the time of cancer diagnosis are paramount for better protein biomarker combination. Although mutations in HGSC patient management. Our findings are therefore BRCA and other homologous recombination genes are an important step in advancing towards a stratified risk- known to predict response to chemotherapy, current management of HGSC patients through the identifica- clinical guidelines recommend referral for genetic test- tion of biological predictors of treatment response. ing at the time of ovarian cancer diagnosis. Hence, since One of the main challenges in this type of projects is the genetic test is not widely available in all clinical set- the selection of samples to be used for the discovery of tings until chemotherapy has commenced, a predictor of protein candidates relevant to our study. In our study we first-line treatment response, at present, cannot include chose tissues for two main reasons. Firstly, archival FFPE the genetic information. Although we have not integrated tumour samples represent a valuable resource for study- the BRCA mutational status in our HGSC-1LTR strategy, ing cancer biomarkers because they are widely avail- it would be a basic parameter to add in future valida- able in hospitals and are associated to important clinical tion phases together with the use of targeted therapies in information (e.g. histology, response to treatment and the maintenance setting given the rapid advances in the outcomes). Secondly, all patients included in the study, field. In addition, our study was limited by the availabil- and for whom the predictive biomarker is intended to be ity of HGSC samples and the quality of the FFPE tissues used (i.e. patients with newly diagnosed advanced stage for protein extraction and MS analysis. This limitation HGSC), undergo a biopsy prior to starting chemotherapy is associated to differences in patients’ age between the treatment. Therefore, patients’ tissues are available with- chemoresistant and chemosenstivie groups and TFIp out the need of additional invasive procedures. Moreover, between the discovery and the verification cohorts. tumour tissues are where potential protein biomarkers Given that we included patients undergoing either pri- are more likely to be enriched. Although FFPE tissues mary cytoreductive surgery or interval surgery, it would have traditionally been associated to a high variation in also be interesting to evaluate the HGSC-1LTR strategy protein quality due to formalin-induced chemical modi- in future studies comparing the molecular characteristics fications and differences in storage times [34], recent of tissues before and after neoadjuvant chemotherapy. studies have confirmed that proteomes are preserved to Finally, for the widespread implementation of our signa- a comparable extent to those obtained from fresh frozen ture in routine clinical practice, we would require valida- tissues and are not influenced by their storage [35, 36]. tion in a higher number of patients and the availability of In the study by Coscia et al. [11] authors evaluated 25 mass spectrometry technology in the clinics. advanced stage HGSC FFPE samples by discovery pro- teomics and identified CT45 as a platinum sensitivity Conclusions mediator in ovarian cancer. We were able to observe an We have established a new strategy (HGSC-1LTR) that increased protein abundance by targeted proteomics in combines tissue levels of proteins TKT, LAMC1 and chemosensitive samples, in line with the results observed FUCO, together with patients’ age, menopausal status, by Coscia et al. However, this protein was excluded from serum CA125 levels, and treatment approach to predict subsequent predictive analysis because its targeted pep- the response to first-line treatment in HGSC patients. tides were not consistently detected. These data are obtained from ready-available biopsies Clinical parameters routinely recorded by healthcare in hospitals, and ready-available clinical data at the time professionals can be relevant to improve the classifica- of HGSC diagnosis, thus no additional interventions tory ability of protein biomarkers. An example of this are required beyond current clinical practice. Because is the Risk of Ovarian Malignancy Algorithm (ROMA) predictive tools are currently lacking for patients with which integrates patients menopausal status to serum advanced HGSC, this new strategy is clinically relevant CA125 and HE4 levels, to distinguishing between a for the prediction of chemoresistant patients. Identifi- benign and malignant pelvic mass [37]. We assessed the cation of chemoresistance at the time of diagnosis can classificatory ability of our protein signature in combina- facilitate the study of alternative treatments aimed at tion with relevant clinical data such as the patients age, improving the outcome for these patients. In addition, menopausal status, serum CA125 levels and the deci- those patients classified as chemosensitive could undergo sion to treat with either primary cytoreductive surgery standard care with platinum-based agents. Therefore, the or neoadjuvant chemotherapy. Indeed, the addition of HGSC-1LTR strategy can allow optimization of thera- the patients data that is currently considered in clini- peutic decision making and individualize HGSC patients’ cal practice, improved the classification power of the care. Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 12 of 13 Abbreviations Funding AUC​: Area under the curve; FFPE: Formalin-fixed paraffin-embedded; HGSC: This work was supported by the PhD4MD collaborative research program High-grade serous ovarian carcinoma; PARP inhibitor: Poly (ADP-ribose) between the Vall d’Hebron Research Institute (VHIR) and the Centre for polymerase inhibitor; PRM: Parallel reaction monitoring; ROC curves: Receiver Genomic Regulation (CRG). It has been supported by grants from the Instituto operating characteristic curves; ROMA: Risk of Ovarian Malignancy Algorithm; Carlos III (PI18/01017), the Miguel Servet Program (CPII18/00027) and the Min‑ TFIp: Treatment-free interval to platinum. isterio de Economía y Competitividad y Fondos FEDER (RTC-2015-3821-1 to AS and CTQ2016-80364-P to ES). This project has also received funding from the European Union’s Horizon 2020 research and innovation program under grant Supplementary Information agreement No 823839 (EPIC-XS).The CRG/UPF Proteomics Unit is part of the The online version contains supplementary material available at https://​doi.​ Spanish Infrastructure for Omics Technologies (ICTS OmicsTech) and it is sup‑ org/​10.​1186/​s12967-​022-​03816-7. ported by “Secretaria d’Universitats i Recerca del Departament d’Economia i Coneixement de la Generalitat de Catalunya” (2017SGR595 and 2017SGR1661). Additional file 1: Table S1. Clinical data of the 21 patients included in the We also acknowledge support of the Spanish Ministry of Science and Innova‑ discovery cohort. tion to the EMBL partnership, the Centro de Excelencia Severo Ochoa and the CERCA Programme / Generalitat de Catalunya. Additional file 2: Table S2. Clinical data of the 63 patients included in the verification cohort. Availability of data and materials Additional file 3: Table S3. Discovery cohort quantitative matrix. The datasets generated and analysed during the current study are available in the ProteomeXchange Consortium via the PRIDE [38] partner repository with Additional file 4: Figure S1. Protein quantitative assay development. the dataset identifier PXD035084, https://​www.​ebi.​ac.​uk/​pride/ (A) Histogram showing the number of proteins quantified in each group. In red are proteins quantified in at least 4 out of 7 patients which were included in the statistical analysis. (B) Peptide identity confirmation Declarations between parallel reaction monitoring (PRM) elution profiles of endog‑ enous peptides and internal standards. Transitions showing interferences Ethical approval and consent to participate were removed (middle panel, marked arrow) and those measurements The study was approved by the Institutional Ethical Board (PRAMI3082015) considered to be under the limit of detection were replaced by the and conducted in accordance with the ethical standards for human experi‑ background value (lower panel). (C) Coefficients of variation (CV) of the 52 mentation established in the Declaration of Helsinki. Informed consent was peptides showing the low technical variability of the assay. obtained from all patients. Additional file 5: Table S4. Parallel reaction monitoring quantitative Consent for publication results. Not applicable. Additional file 6: Table S5. Quantitative values (log2 areas) per protein and patient used in the prediction analysis. Competing interests Additional file 7: Table S6. Performance of the predicted proteins and The authors declare that they have no competing interests. their ability to classify patients into chemoresistant and chemosensitive groups. Author details 1 Centre de Regulació Genòmica, Barcelona Institute of Science and Technol‑ Additional file 8: Figure S2. Prediction ability of the protein signatures ogy (BIST), Dr Aiguader 88, 08003 Barcelona, Spain. 2 Universitat Pompeu when considering partially chemosensitive patients and clinical variables. Fabra, Dr Aiguader 88, 08003 Barcelona, Spain. 3 Biomedical Research Group (A) Retention time drift of TKT, LAMC1, FUCO and CRP endogenous in Gynecology, Vall d’Hebron Institut de Recerca, Universitat Autònoma de peptides for all samples analysed. (B) Receiver operating curves of the Barcelona, Vall, d’Hebron Barcelona Hospital Campus, Passeig Vall d’Hebron two best protein combination. TKT + LAMC1 + FUCO with an AUC of 0.76 119‑129, 08035 Barcelona, Spain. 4 Department of Gynecology, Hospital (95% CI 0.64—0.87) and TKT + LAMC1 + CRP with an AUC of 0.75 (95% CI Universitari Vall d’Hebron, Vall d’Hebron Barcelona Hospital Campus, Passeig 0.64–0.86). (C) Receiver operating curves of the two best protein combi‑ Vall d’Hebron 119‑129, 08035 Barcelona, Spain. 5 Department of Pathology, nation classifiers and their ability to discriminate between chemoresistant Hospital Universitari Vall d’Hebron, Vall d’Hebron Barcelona Hospital Campus, and partially chemosensitive patients. TKT + LAMC1 + FUCO with an AUC Passeig Vall d’Hebron 119‑129, 08035 Barcelona, Spain. 6 Centro de Investi‑ of 0.76 (95%CI 0.61–0.91) and TKT + LAMC1 + CRP with an AUC value of gación Biomédica en Red (CIBERONC), Instituto de Salud Carlos III, Avenida de 0.70 (95% CI 0.53–0.87). Monforte de Lemos 3‑5, 28029 Madrid, Spain. 7 Cell Cycle and Cancer Labora‑ Additional file 9: Table S7. Performance of the clinical data included in tory, Biomedical Research Group in Urology, Vall Hebron Institut de Recerca, the exploratory analysis and their ability to classify patients into chemore‑ Vall d’Hebron Barcelona Hospital Campus, Universitat Autònoma de Barcelona, sistant and chemosensitive groups. Passeig Vall d’Hebron 119‑129, 08035 Barcelona, Spain. Received: 30 August 2022 Accepted: 7 December 2022 Acknowledgements The authors are grateful to the team members of the Proteomics Unit at the Centre for Genomic Regulation, the Biomedical Research Group in Gynecol‑ ogy at the Vall d’Hebron Institute, the Gynecological Oncology Unit at the Vall d’Hebron Hospital and the Biomedical Research Group in Urology at the Vall References d’Hebron Institute for their assistance. 1. Coburn SB, Bray F, Sherman ME, Trabert B. International patterns and trends in ovarian cancer incidence, overall and by histologic subtype. Int Author contributions J cancer. 2017;140(11):2451–60. Conceptualization: M.B, E.S, A.S. Methodology: M.B, E.B, E.S, A.S. Data curation, 2. Colombo N, Sessa C, Du Bois A, Ledermann J, McCluggage WG, McNeish experiments and analysis: M.B, E.B, M.V. Data interpretation: M.B, E.B, E.S, A.S. I, et al. ESMO-ESGO consensus conference recommendations on ovarian Provision of resources: M.B, J.C, A.G, A.P, J.S, E.S, A.S. Writing – Original draft: cancer: pathology and molecular biology, early and advanced stages, M.B, E.B, E.S, A.S. Review – Original draft: M.B, E.B, M.V, J.C, A.G, A.P, J.S, E.S, borderline tumours and recurrent disease†. Ann Oncol Off J Eur Soc Med A.S. Supervision: E.S, A.S. Funding acquisition: E.S, A.S. All authors read and Oncol. 2019;30(5):672–705. approved the final manuscript. 3. Ray-Coquard I, Pautier P, Pignata S, Pérol D, González-Martín A, Berger R, et al. Olaparib plus bevacizumab as first-line maintenance in ovarian cancer. N Engl J Med. 2019;381(25):2416–28. Bradbury et al. Journal of Translational Medicine (2022) 20:611 Page 13 of 13 4. Coleman RL, Fleming GF, Brady MF, Swisher EM, Steffensen KD, Fried‑ 27. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, et al. pROC: lander M, et al. Veliparib with first-line chemotherapy and as maintenance an open-source package for R and S+ to analyze and compare ROC therapy in ovarian cancer. N Engl J Med. 2019;381(25):2403–15. curves. BMC Bioinform. 2011;17:12. 5. Colombo PE, Fabbro M, Theillet C, Bibeau F, Rouanet P, Ray-Coquard I. 28. Wilson MK, Pujade-Lauraine E, Aoki D, Mirza MR, Lorusso D, Oza AM, et al. Sensitivity and resistance to treatment in the primary management of Fifth Ovarian Cancer Consensus Conference of the Gynecologic Cancer epithelial ovarian cancer. Crit Rev Oncol Hematol. 2014;89(2):207–16. InterGroup: recurrent disease. Ann Oncol Off J Eur Soc Med Oncol. 6. Tomao F, D’Incalci M, Biagioli E, Peccatori FA, Colombo N. Restoring plati‑ 2017;28(4):727–32. num sensitivity in recurrent ovarian cancer by extending the platinum- 29. Rifai N, Gillette MA, Carr SA. Protein biomarker discovery and valida‑ free interval: myth or reality? Cancer. 2017;123(18):3450–9. tion: the long and uncertain path to clinical utility. Nat Biotechnol. 7. Verhaak RGW, Tamayo P, Yang JY, Hubbard D, Zhang H, Creighton CJ, et al. 2006;24(8):971–83. Prognostically relevant gene signatures of high-grade serous ovarian 30. Chiva C, Sabidó E. Peptide selection for targeted protein quantitation. J carcinoma. J Clin Invest. 2013;123(1):517–25. Proteome Res. 2017;16(3):1376–80. 8. Huang D, Chowdhury S, Wang H, Savage SR, Ivey RG, Kennedy JJ, et al. 31. Pignata S, Scambia G, Bologna A, Signoriello S, Vergote IB, Wagner U, Multiomic analysis identifies CPT1A as a potential therapeutic target in et al. Randomized controlled trial testing the efficacy of platinum-free platinum-refractory, high-grade serous ovarian cancer. Cell Rep Med. interval prolongation in advanced ovarian cancer: the MITO-8, MaNGO, 2021;2(12). BGOG-Ov1, AGO-Ovar2.16, ENGOT-Ov1, GCIG Study. J Clin Oncol. 9. Norquist B, Wurz KA, Pennil CC, Garcia R, Gross J, Sakai W, et al. Secondary 2017;35(29):3347–53. somatic mutations restoring BRCA1/2 predict chemotherapy resistance 32. Dockery LE, Rubenstein AR, Ding K, Mashburn SG, Burkett WC, Davis AM, in hereditary ovarian carcinomas. J Clin Oncol. 2011;29(22):3008–15. et al. Extending the platinum-free interval: the impact of omitting 2nd 10. Patch AM, Christie EL, Etemadmoghadam D, Garsed DW, George J, Fere‑ line platinum chemotherapy in intermediate platinum-sensitive ovarian day S, et al. Whole-genome characterization of chemoresistant ovarian cancer. Gynecol Oncol. 2019;155(2):201–6. cancer. Nature. 2015;521(7553):489–94. 33. Pujade-Lauraine E, Banerjee S, Pignata S. Management of platinum- 11. Coscia F, Lengyel E, Duraiswamy J, Ashcroft B, Bassani-Sternberg M, resistant, relapsed epithelial ovarian cancer and new drug perspectives. J Wierer M, et al. Multi-level proteomics identifies CT45 as a chemosen‑ Clin Oncol. 2019;37(27):2437–48. sitivity mediator and immunotherapy target in ovarian cancer. Cell. 34. Gaffney EF, Riegman PH, Grizzle WE, Watson PH. Factors that drive the 2018;175(1):159-170.e16. increasing use of FFPE tissue in basic and translational cancer research. 12. Kim SI, Jung M, Dan K, Lee S, Lee C, Kim HS, et al. Proteomic discovery of Biotech Histochem. 2018;93(5):373–86. biomarkers to predict prognosis of high-grade serous ovarian carcinoma. 35. Zhu Y, Weiss T, Zhang Q, Sun R, Wang B, Yi X, et al. High-throughput Cancers (Basel). 2020;12(4):790. proteomic analysis of FFPE tissue samples facilitates tumor stratification. 13. Bradbury M, Borràs E, Pérez-Benavente A, Gil-Moreno A, Santamaria A, Mol Oncol. 2019;13(11):2305–28. Sabidó E. Proteomic Studies on the management of high-grade serous 36. Piehowski PD, Petyuk VA, Sontag RL, Gritsenko MA, Weitz KK, Fillmore ovarian cancer patients: a mini-review. Cancers (Basel). 2021;13(9):2067. TL, et al. Residual tissue repositories as a resource for population-based 14. Geyer PE, Holdt LM, Teupser D, Mann M. Revisiting biomarker discovery cancer proteomic studies. Clin Proteomics. 2018. https://​doi.​org/​10.​1186/​ by plasma proteomics. Mol Syst Biol. 2017;13(9):942. s12014-​018-​9202-4. 15. Whiteaker JR, Lin C, Kennedy J, Hou L, Trute M, Sokal I, et al. A targeted 37. Moore RG, Miller MC, Disilvestro P, Landrum LM, Gajewski W, Ball JJ, et al. proteomics-based pipeline for verification of biomarkers in plasma. Nat Evaluation of the diagnostic accuracy of the risk of ovarian malignancy Biotechnol. 2011;29(7):625–34. algorithm in women with a pelvic mass. Obstet Gynecol. 2011;118(2 Pt 16. Bradbury M, Borràs E, Castellví J, Méndez O, Sánchez-Iglesias JL, Pérez- 1):280–8. Benavente A, et al. BRCA1 mutations in high-grade serous ovarian cancer 38. Perez-Riverol Y, Bai J, Bandla C, García-Seisdedos D, Hewapathirana S, are associated with proteomic changes in DNA repair, splicing, transcrip‑ Kamatchinathan S, et al. The PRIDE database resources in 2022: a hub tion regulation and signaling. Sci Rep. 2022;12(1):625–34. for mass spectrometry-based proteomics evidences. Nucleic Acids Res. 17. Gámez-Pozo A, Trilla-Fuertes L, Prado-Vázquez G, Chiva C, López-Vacas 2022;50(D1):D543–52. R, Nanni P, et al. Prediction of adjuvant chemotherapy response in triple negative breast cancer with discovery and targeted proteomics. PLoS One. 2017;12(6). Publisher’s Note 18. Beer LA, Liu P, Ky B, Barnhart KT, Speicher DW. Efficient quantitative com‑ Springer Nature remains neutral with regard to jurisdictional claims in pub‑ parisons of plasma proteomes using label-free analysis with MaxQuant. lished maps and institutional affiliations. Methods Mol Biol. 2017;1619:339–52. 19. Choi M, Chang CY, Clough T, Broudy D, Killeen T, MacLean B, et al. MSstats: an R package for statistical analysis of quantitative mass spectrometry- based proteomic experiments. Bioinformatics. 2014;30(17):2524–6. 20. Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A. 2003;100(16):9440–5. 21. Kleinberg L, Holth A, Trope CG, Reich R, Davidson B. Claudin upregula‑ tion in ovarian carcinoma effusions is associated with poor survival. Hum Pathol. 2008;39(5):747–57. 22. Yang W, Cho H, Shin HY, Chung JY, Kang ES, Lee E, et al. Accumulation of Ready to submit your research ? Choose BMC and benefit from: cytoplasmic Cdk1 is associated with cancer growth and survival rate in epithelial ovarian cancer. Oncotarget. 2016;7(31):49481–97. • fast, convenient online submission 23. Liu Y, Chen S, Zheng C, Ding M, Zhang L, Wang L, et al. The prognostic value of the preoperative c-reactive protein/albumin ratio in ovarian • thorough peer review by experienced researchers in your field cancer. BMC Cancer. 2017. https://​doi.​org/​10.​1186/​s12885-​017-​3220-x. • rapid publication on acceptance 24. Chiva C, Olivella R, Borràs E, Espadas G, Pastor O, Solé A, et al. QCloud: A • support for research data, including large and complex data types cloud-based quality control system for mass spectrometry-based prot‑ eomics laboratories. PLoS One. 2018;13(1):e0189209. • gold Open Access which fosters wider collaboration and increased citations 25. MacLean B, Tomazela DM, Shulman N, Chambers M, Finney GL, Frewen B, • maximum visibility for your research: over 100M website views per year et al. Skyline: an open source document editor for creating and analyzing targeted proteomics experiments. Bioinformatics. 2010;26(7):966–8. At BMC, research is always in progress. 26. Borràs E, Cantó E, Choi M, Villar LM, Álvarez-Cermeño JC, Chiva C, et al. Protein-based classifier to predict conversion from clinically isolated Learn more biomedcentral.com/submissions syndrome to multiple sclerosis. Mol Cell Proteomics. 2016;15(1):318–28.