Published online 21 October 2020 Nucleic Acids Research, 2021, Vol. 49, Database issue D687–D693 doi: 10.1093/nar/gkaa890 DualSeqDB: the host–pathogen dual RNA sequencing database for infection processes Javier Macho Rendón1,† , Benjamin Lang 2,3,† , Marc Ramos Llorens1 , Gian Gaetano Tartaglia 2,4,5,* and Marc Torrent Burgas 1,* 1 Systems Biology of Infection Lab, Department of Biochemistry and Molecular Biology, Biosciences Faculty, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain, 2 Centre for Genomic Regulation (CRG), The Barcelona Institute for Science and Technology, Dr. Aiguader 88, 08003 Barcelona, Spain, 3 Department of Structural Biology, St. Jude Children’s Research Hospital, Memphis, TN 38105, USA, 4 Department of Neuroscience and Brain Technologies, Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genoa, Italy and 5 Department of Biology ‘Charles Darwin’, Sapienza University of Rome, Ple A. Moro 5, 00185 Rome, Italy Received August 12, 2020; Revised September 27, 2020; Editorial Decision September 28, 2020; Accepted September 29, 2020 ABSTRACT INTRODUCTION Despite antibiotic resistance being a matter of grow- During infection, pathogens trigger the expression of ing concern worldwide, the bacterial mechanisms unique genes that ensure their survival and allow replicating of pathogenesis remain underexplored, restraining within the host. In turn, the host activates complex mecha- our ability to develop new antimicrobials. The rise of nisms to recognize and kill pathogens. Hence, the simulta- high-throughput sequencing technology has made neous detection of host and pathogen transcripts during the infection process can provide deeper insights into the host– available a massive amount of transcriptomic data pathogen interaction than those detected from the host or that could help elucidate the mechanisms underly- pathogen in isolation. The term ‘dual RNA-seq’ refers to ing bacterial infection. Here, we introduce the Du- the process of simultaneously analyzing RNA-seq data of a alSeqDB database, a resource that helps the iden- pathogenic bacteria and the infected host (1). Dual RNA- tification of gene transcriptional changes in both seq has become a leading approach to uncover the intri- pathogenic bacteria and their natural hosts upon in- cate relationship between pathogen and host interactions al- fection. DualSeqDB comprises nearly 300 000 entries lowing researchers to identify ‘molecular phenotypes’ that from eight different studies, with information on bac- would otherwise remain undetected (2–4). terial and host differential gene expression under in In a typical dual RNA-seq experiment, either animals are vivo and in vitro conditions. Expression data values inoculated with a defined load of bacteria (in vivo) or rel- evant cell culture models are inoculated with bacteria at a were calculated entirely from raw data and analyzed defined multiplicity of infection (in vitro). After inoculation, through a standardized pipeline to ensure consis- samples are taken over time to determine the time response. tency between different studies. It includes informa- At each time point, infected cells are lysed, RNA is iso- tion on seven different strains of pathogenic bacteria lated and the cDNA library is prepared and sequenced us- and a variety of cell types and tissues in Homo sapi- ing high-throughput sequencing technologies, which gener- ens, Mus musculus and Macaca fascicularis at dif- ates large amounts of data. RNA-seq data of mock-infected ferent time points. We envisage that DualSeqDB can host cells and initial bacterial cultures are used as control help the research community in the systematic char- conditions for expression analysis. Dual RNA-seq experi- acterization of genes involved in host infection and ments have several technical difficulties, including the dif- help the development and tailoring of new molecules ferent nature and content of RNA between bacteria and against infectious diseases. DualSeqDB is freely eukaryotic cells, the larger proportion of RNA from eu- karyotic cells and the need to account for the prevalence of available at http://www.tartaglialab.com/dualseq. rRNA transcripts and variable infection rates (1,5). Usually, such limitations can be solved using high-depth sequenc- ing, pathogen and host rRNA depletion, and enrichment * To whom correspondence should be addressed. Tel: +34 93 581 4147; Email: marc.torrent@uab.cat Correspondence may also be addressed to Gian Gaetano Tartaglia. Tel: +34 93 316 0116; Email: gian.tartaglia@crg.eu † The authors wish it to be known that, in their opinion, the first two authors should be regarded as Joint First Authors. C The Author(s) 2020. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com D688 Nucleic Acids Research, 2021, Vol. 49, Database issue Table 1. List of dual RNA-seq studies included in DualSeqDB Pathogen Host organism Tissue/cell type Condition GEO code Reference Streptococcus pyogenes Macaca fascicularis Skeletal muscle tissue In vivo GSE144100 (9) Salmonella enterica serovar Homo sapiens HeLa-S3 cells In vitro GSE60144 (4) Typhimurium SL1344 Salmonella enterica serovar Homo sapiens Endothelial cells In vitro GSE136717 (10) Typhimurium SL1344 Epithelial cells Monocytic cells NK cells Yersinia pseudotuberculosis IP 32953 Mus musculus Lymphoid tissue In vivo PRJEB14242 (ENA) (3) Pseudomonas aeruginosa PA01 Mus musculus Lung tissue In vivo SRP090213 (SRA) (11) Haemophilus ducreyi 35000HP Homo sapiens Skin tissue In vivo GSE130901 (12) Mycobacterium tuberculosis ATCC Homo sapiens THP-1 cells In vitro PRJEB6552 (ENA) (13) 35733 Streptococcus pneumoniae D39 Homo sapiens A549 cells In vitro GSE79595 (2) of samples for infected host cells by fluorescence-activated were mapped to host genome index with HISAT2. Mapped cell sorting (6). reads were stored as BAM files, and unmapped reads were Dual RNA-seq is a mixture of host and pathogen tran- kept in a separate FastQ file. FeatureCounts (18), together scripts where different RNA samples may contain variable with the host annotation file, was used for gene counting, proportions of pathogen to host reads (7,8). These tran- and a matrix of read counts was generated where each row scripts can be sorted into the corresponding organisms by represents an annotated gene and each column represents a different computational strategies (1,2) and the accuracy different condition or biological replicate. Unmapped reads of differential expression values depends on the analytical from the previous mapping step were then mapped back to method used. To circumvent these biases, we need a stan- the bacterial genome index with Bowtie2, and a matrix of dard pipeline to compare data from different sources. read counts was produced similarly by using the bacterial Despite the increasing availability of raw sequencing data annotation file and FeatureCounts. HISAT2 and Bowtie2 from dual RNA-seq experiments, the existence of multi- are run with default parameters as a way to simplify and ple analysis pipelines may hinder the comparison between standardize criteria when analyzing data coming from het- datasets. Dual RNA-seq pipelines are very sensitive to soft- erogeneous sources. Finally, to calculate gene expression ware selection and parameter definition. This lack of stan- changes in treated against control conditions, differential dardization motivated us to create DualSeqDB, a user- expression analysis was performed separately for the bac- friendly platform to search for changes in gene expression terial and the host matrices by using the DESeq2 R Pack- levels during infection at both pathogen and host levels. age (19). For this, the Wald test was used under the null hy- To build this database, we analyzed raw sequencing data pothesis that there is no differential expression between the from heterogeneous dual RNA-seq studies using a well- control and the treated samples. The estimated gene expres- defined pipeline, to generate comparable gene expression sion change value [measured in log2 fold change (FC)] and data. This setup allows DualSeqDB to compare across mul- its associated P-value were generated for each annotated tiple species and experimental conditions. gene with detected reads in at least one condition (Figures 1 and 2). P-values were corrected for multiple testing using the Benjamini–Hochberg method. All additional informa- MATERIALS AND METHODS tion such as bacterial ID, host ID, time point, experimental Processing sequencing data condition (in vivo/in vitro) and cell type/tissue was added to each gene to create the final format as displayed in DualSe- To build DualSeqDB, we reprocessed raw data from avail- qDB. able studies (Table 1) and used a well-defined pipeline to provide robust and homogeneous information in our Technical aspects database (Figure 1). To this end, we selected only dual RNA-seq studies containing at least two biological repli- DualSeqDB was built using PHP on an Apache web server cates and only when data were available for infected and with a MySQL database backend. Sequence identifiers and control conditions for both pathogen and host (2–4,9–13. cross-references were obtained from UniProt and the NCBI For each study, genome and annotation files were down- RefSeq, Gene and Genome resources (20). DualSeqDB loaded from the NCBI Reference Sequence Database (Ref- stores no user data, except for the anonymous caching of Seq) (14). Bacterial and eukaryotic genome indices were BLAST search results for a given sequence in order to created with Bowtie2 (15) and HISAT2 (16), respectively. greatly speed up repeated searches. The open-source Boot- HISAT2 can take into account alternative splicing of genes strap library was used to allow display on devices of any and was used for eukaryotic genome indexing. For each bi- screen size, including mobile devices. Several icons were in- ological replicate, raw sequencing reads in FastQ format cluded from Font Awesome and the Noun Project, and a were trimmed with Trimmomatic (17) to remove adapter number of JavaScript libraries are used for table export and content. During this process, reads that are <36 bases long sorting. Please see the About section of the website for de- are dropped from the analysis. Afterward, surviving reads tailed attributions. Nucleic Acids Research, 2021, Vol. 49, Database issue D689 Figure 1. Pipeline used to process raw sequencing data from dual RNA-seq studies. Raw sequencing data were downloaded from the corresponding repository in FastQ format and adapter sequences were removed with Trimmomatic. Pathogen and host genomes and annotations were downloaded to build genome indices. Trimmed FastQ files were then mapped to the host index genome with HISAT2 and the unmapped reads were subsequently mapped to the pathogen index genome using Bowtie2. From this point onward, pathogen and host reads were analyzed in parallel. Mapped reads were quantified with FeatureCounts and their respective annotation files, creating a matrix of read counts. This matrix containing control and treated samples is then used as input for the DESeq2 R package to perform a differential expression analysis. The differential gene expression changes (measured as log2 FC) and corresponding P-values (Benjamini–Hochberg correction for multiple testing) were calculated using DESeq2. D690 Nucleic Acids Research, 2021, Vol. 49, Database issue Figure 2. Visualization of overall statistical significance (P-value) and magnitude of change (log2 FC) of all entries in DualSeqDB. Gene expression changes were considered significant when log2 FC > 1 (upregulated) or log2 FC < −1 (downregulated) when the corresponding P-value was <0.05 (dashed blue lines). Pathogen genes are labelled in yellow and host genes in green. BLAST search identifiers, UniProt protein accessions or a free-text search in the gene product’s description (Figure 3). To search The NCBI BLAST suite version 2.9.0+ (March 2019) (21) within a particular host and/or pathogen, users can select is used to search by sequence similarity. The BLASTP pro- the pathogen and/or host name in the drop-down menu. If gram is used for amino acid sequences, and BLASTX for no gene or protein name is given, the output will display a nucleic acid (coding) sequences. BLAST search results are complete list of genes, similar to the Browse view (described cached for each unique sequence, which means that re- below). running a search using the same sequence will yield results nearly instantaneously. As on all other pages, results from the BLAST search page can be linked to and shared with Tables on DualSeqDB: sorting, downloading and linking to other researchers using the ‘Link to these results’ link at the results bottom of the page. For sequences above a URL length of After selecting a gene of interest, a view will open with all 2000 characters, this link uses a sequence hash identifying the information available for the corresponding gene (Fig- the cached sequence, rather than the sequence itself. ure 4). The heading of this page provides information on the selected protein: protein and host/pathogen name, length, Protein visualization gene name and UniProt ID. In the table, all available ex- perimental data are listed: tissue of the host organism, tis- For UniProt proteins, a protein visualization is automat- sue condition (whether the experiment was carried out in ically generated and displayed by ProViz (22). ProViz is vivo or in vitro), time after infection, differential expres- an interactive exploration tool for investigating the struc- sion gene data, including the log2 FC and the associated tural, functional and evolutionary features of proteins and P-value, a note giving information on the growth condi- is likely to be particularly helpful for analyzing uncharac- tions of control bacteria (including temperature and growth terized proteins. phase, whenever specified in its study, otherwise it is shown as ‘none’) and the reference to the original paper where the USING DUALSEQDB data were published. A brief description on the meaning of log2 FC and P- DualSeqDB consists of several elements: (i) a text search value is also available as mouse-over explanation on the function to find specific eukaryotic and bacterial genes; (ii) column headers. For any proteins in UniProt, a protein vi- a BLAST search function to find genes similar to a protein sualization is automatically provided by ProViz from the or nucleic acid sequence of interest; (iii) a Browse function Davey lab. Proteins >5000 amino acids are not displayed to quickly identify genes up- and downregulated during in- due to display speed limitations. ProViz is an interactive ex- fection; and (iv) a Tutorial section to get started quickly ploration tool that allows inspection of the structural, func- by following a step-by-step guide. DualSeqDB relies on tional and evolutionary features of proteins, including Pfam JavaScript; therefore, users will need to enable this in their domains and transmembrane regions. This tool is particu- web browser for full functionality. larly useful for unknown and uncharacterized proteins. Alternatively, the protein’s FASTA sequence can be dis- played by pressing the ‘Show protein sequence’ button, Search function along with a ‘Copy’ link in the top right corner to copy To search for a gene or protein, users simply need to type its and paste the protein’s sequence into other research tools, name or identifier. Any of the following options are avail- or into the DualSeqDB BLAST Search to search for sim- able: gene symbols, gene locus identifiers, NCBI protein ilar proteins. You can also immediately search for similar Nucleic Acids Research, 2021, Vol. 49, Database issue D691 Figure 3. Search results example summary. The search results page displays a list of any host or bacterial genes matching the search term. It also displays information on the infected host species and its associated pathogenic bacteria, the NCBI protein identifier, the UniProt protein accession code and the gene symbol of the gene for which the expression change was measured. This preview section also shows a description of the gene product and its length, together with the expression change value (measured as log2 FC) and the corresponding P-value. In this example, we show the case of CXCL2, a chemoattractant chemokine with pro-inflammatory function, involved in many immune responses, such as cancer metastasis, wound healing or angiogenesis. The results collected in DualSeqDB show that, upon infection of skin tissue human cells with the pathogen Haemophilus ducreyi, the human gene CXCL2 increases its expression levels, as indicated by log2 FC > 4. Figure 4. Detailed view of gene expression changes. The detailed view page displays all the information available for a host or bacterial gene, along with a ProViz visualization of the protein’s sequence and structural features, showing sequence conservation with similar proteins. It also shows the log2 FC and the associated P-value for each entry, together with all the details of the experiment: host name, pathogen name, organism (indicating whether the measured gene belongs to the host or the pathogen), cell type/tissues, post-infection time points, as well as the PMID reference with a PubMed link to the original study, and a note column, specifying the bacterial growth conditions if listed in the original study. proteins via BLAST (see below for more details) by press- BLAST search ing the ‘Find similar proteins’ button. The BLAST Search tab provides a search by sequence simi- To sort the table as desired, users can select any of the larity. When the protein of interest is not in our database, the column headers. The current table can be downloaded as a user may search for similar proteins using BLAST sequence comma-separated CSV file for export into spreadsheet soft- alignment. Finding a similar protein with a high variation ware such as Microsoft Excel using the ‘Download Table’ in log2 FC and low P-value is a strong indication that the button in the top right. An appropriate readable file name query sequence may be relevant during infection. To search is automatically generated. The results can also be linked for similar proteins in our database using BLAST, protein to and shared with other researchers by right-clicking and or coding sequences in FASTA format can be used and have copying the ‘Link to these results’ link at the bottom of the to be properly identified in the drop-down menu. When the page. D692 Nucleic Acids Research, 2021, Vol. 49, Database issue BLAST alignment is ready, a search results page will open DISCUSSION with the following information: The development of new antimicrobial therapies heavily re- lies on our knowledge of the mechanisms of bacterial in- 1. Identity: The percentage of sequence identity between fection (23–25). Therefore, it is crucial to understand how query and target in the successfully aligned region. bacterial infection develops and which bacterial genes are 2. Aligned: The total number of amino acids that were suc- required to infect a host . The use of high-throughput se- cessfully aligned between query and target. quencing technologies has unveiled new levels of complex- 3. Bit score: The required size of a sequence database in ity in the transcriptomic response of pathogens and hosts which the current match could be found just by chance. during infection. In the last few years, dual RNA-seq has The bit score is a log2 -scaled and normalized raw score, become the leading approach to uncover the intricate re- meaning that each increase by one doubles the required lationship between pathogen and host interactions. Hence, database size. dual RNA-seq could be used to define host–pathogen in- 4. E-value: The number of expected hits of similar qual- teractions or identify potential biomarkers of infection (26). ity (score) that could be found in the BLAST sequence At present, dual RNA-seq data are disseminated in multiple database just by chance. locations and incompatible formats and are therefore not The meaning of the Host, Pathogen, Locus, Protein, accessible to the scientific community without specialized Gene, Product, P-value and log2 FC columns can be found tools and knowledge. in the Browse tab section below, or via the mouse-over DualSeqDB intends to be a valuable central resource for information symbols in the top row of any table. By de- the systematic identification of proteins that are crucial for fault, BLAST matches with the highest bit scores are shown successful infection, aimed to understand how the bacte- first and matches with 100% sequence identity will be high- rial and host transcriptomes change and interact during in- lighted in green. Tables can also be sorted as desired us- fection. In this context, we envisage that DualSeqDB will ing the column headers. As for all tables, the results can facilitate the finding of interspecies relationships between be downloaded as a comma-separated CSV file for export pathogen and host and will help us to uncover new mech- into spreadsheet software such as Microsoft Excel using the anisms of infection. The analysis of the results included in ‘Download Table’ button in the top right corner. An appro- DualSeqDB may inspire the design of new therapeutic inter- priate readable file name is automatically generated. The re- ventions aimed to prevent the spread of infection. Given the sults can also be linked to and shared with other researchers current momentum of sequencing technologies in research by right-clicking and copying the ‘Link to these results’ link and clinics, we expect that our database will grow continu- at the bottom of the results table. ously and become a comprehensive repository that will help us in the fight against infectious diseases. Browsing the entire database DATA AVAILABILITY The Browse tab provides an overview of all entries in the To download the entire DualSeqDB database for local anal- DualSeqDB database. A pathogenic species or a host of ysis, please click the link available under the Download tab. interest can be chosen in the selection element at the top. Currently, DualSeqDB v1 is available, and will be upgraded This table is sorted by significance and log2 FC. It displays with new data as they become available. pathogen/host genes with a high and significant change in expression during infection at the top, followed by insignif- icant genes by decreasing log2 FC in absolute value. Genes ACKNOWLEDGEMENTS with very little expression changes and high P-value are listed at the very end of the table. Arrows next to each field The authors would like to thank Dr Norman E. Davey for provide links to useful external databases: his kind support in providing ProViz for protein visualiza- tion. 1. Pathogen/Host: Links out to the NCBI Taxonomy database, a comprehensive taxonomic database. FUNDING 2. Locus: Links out to the Ensembl database, which pro- vides genome annotation for all species included in the Ministerio de Ciencia, Innovación y Universidades database. [SAF2015-72518-EXP, SAF2017-82158-R, BFU2017- 3. Protein: Links out to the NCBI Protein database, which 86970-P and RYC-2012-09999]; European Research Coun- provides protein sequences and information. cil [RIBOMYLOME 309545]; Horizon 2020––Research 4. UniProt Accession and Gene Symbol: Links out to the and Innovation Framework Programme [727658, 793135 UniProt Knowledgebase, which provides comprehensive to B.L.]; Ministry of Economy and Competitiveness protein annotation. [BFU2017-86970-P]; Centro de Excelencia Severo Ochoa 2013–2017; Generalitat de Catalunya; European Society of Users can select the Locus, Protein, and UniProt Acces- Clinical Microbiology and Infectious Diseases Research sion and Gene Symbol entries to view details for the given Grant 2016. Funding for open access charge: Ministerio de protein in the external databases. This information is also Ciencia, Innovación y Universidades [BFU2017-86970-P]. available as a mouse-over explanation in the Browse tab. Conflict of interest statement. None declared. Nucleic Acids Research, 2021, Vol. 49, Database issue D693 REFERENCES 13. Rienksma,R.A., Suarez-Diez,M., Mollenkopf,H.J., Dolganov,G.M., Dorhoi,A., Schoolnik,G.K., Martins Dos Santos,V.A., 1. Westermann,A.J., Gorski,S.A. and Vogel,J. (2012) Dual RNA-seq of Kaufmann,S.H., Schaap,P.J. and Gengenbacher,M. (2015) pathogen and host. Nat. Rev. Microbiol., 10, 618–630. Comprehensive insights into transcriptional adaptation of 2. Aprianto,R., Slager,J., Holsappel,S. and Veening,J.W. (2016) intracellular mycobacteria by microbe-enriched dual RNA Time-resolved dual RNA-seq reveals extensive rewiring of lung sequencing. BMC Genomics, 16, 34. epithelial and pneumococcal transcriptomes during early infection. 14. O’Leary,N.A., Wright,M.W., Brister,J.R., Ciufo,S., Haddad,D., Genome Biol., 17, 198. McVeigh,R., Rajput,B., Robbertse,B., Smith-White,B., Ako-Adjei,D. 3. Nuss,A.M., Beckstette,M., Pimenova,M., Schmuhl,C., Opitz,W., et al. (2016) Reference sequence (RefSeq) database at NCBI: current Pisano,F., Heroven,A.K. and Dersch,P. (2017) Tissue dual RNA-seq status, taxonomic expansion, and functional annotation. Nucleic allows fast discovery of infection-specific functions and Acids Res., 44, D733–D745. riboregulators shaping host–pathogen transcriptomes. Proc. Natl 15. Langmead,B. and Salzberg,S.L. (2012) Fast gapped-read alignment Acad. Sci. U.S.A., 114, E791–E800. with Bowtie 2. Nat. Methods, 9, 357–359. 4. Westermann,A.J., Forstner,K.U., Amman,F., Barquist,L., Chao,Y., 16. Kim,D., Paggi,J.M., Park,C., Bennett,C. and Salzberg,S.L. (2019) Schulte,L.N., Muller,L., Reinhardt,R., Stadler,P.F. and Vogel,J. Graph-based genome alignment and genotyping with HISAT2 and (2016) Dual RNA-seq unveils noncoding RNA functions in HISAT-genotype. Nat. Biotechnol., 37, 907–915. host–pathogen interactions. Nature, 529, 496–501. 17. Bolger,A.M., Lohse,M. and Usadel,B. (2014) Trimmomatic: a flexible 5. Westermann,A.J., Barquist,L. and Vogel,J. (2017) Resolving trimmer for Illumina sequence data. Bioinformatics, 30, 2114–2120. host–pathogen interactions by dual RNA-seq. PLoS Pathog., 13, 18. Liao,Y., Smyth,G.K. and Shi,W. (2019) The R package Rsubread is e1006033. easier, faster, cheaper and better for alignment and quantification of 6. Saliba,A.E., Cantos,S.C. and Vogel,J. (2017) New RNA-seq RNA sequencing reads. Nucleic Acids Res., 47, e47. approaches for the study of bacterial pathogens. Curr. Opin. 19. Love,M.I., Huber,W. and Anders,S. (2014) Moderated estimation of Microbiol., 35, 78–87. fold change and dispersion for RNA-seq data with DESeq2. Genome 7. Baddal,B., Muzzi,A., Censini,S., Calogero,R., Torricelli,G., Biol., 15, 550. Guidotti,S. and Paxxicoli,A. (2015). Dual RNA-seq of nontypeable 20. NCBI,Resource Coordinators (2016) Database resources of the Haemophilus influenzae and host cell transcriptomes reveals novel National Center for Biotechnology Information. Nucleic Acids Res., insights into host–pathogen cross talk. mBio, 6, e0176515. 44, D7–D19. 8. Choi,Y.-J., Aliota,M. T., Mayhew,G. F., Erickson,S. M. and 21. Camacho,C., Coulouris,G., Avagyan,V., Ma,N., Papadopoulos,J., Christensen,B. M. (2014). Dual RNA-seq of parasite and host reveals Bealer,K. and Madden,T.L. (2009) BLAST+: architecture and gene expression dynamics during filarial worm–mosquito applications. BMC Bioinformatics, 10, 421. interactions. PLOS Negl. Trop. Dis., 8, e2905. 22. Jehl,P., Manguy,J., Shields,D.C., Higgins,D.G. and Davey,N.E. 9. Kachroo,P., Eraso,J.M., Olsen,R.J., Zhu,L., Kubiak,S.L., Pruitt,L., (2016) ProViz: a web-based visualization tool to investigate the Yerramilli,P., Cantu,C.C., Ojeda Saavedra,M., Pensar,J. et al. (2020) functional and evolutionary features of protein sequences. Nucleic New pathogenesis mechanisms and translational leads identified by Acids Res., 44, W11–W15. multidimensional analysis of necrotizing myositis in primates. mBio, 23. Crua Asensio,N., Muñoz Giner,E., de Groot,N.S. and Torrent 11, e03363-19. Burgas,M. (2017) Centrality in the host–pathogen interactome is 10. Schulte,L.N., Schweinlin,M., Westermann,A.J., Janga,H., associated with pathogen fitness during infection. Nat. Commun., 8, Santos,S.C., Appenzeller,S., Walles,H., Vogel,J. and Metzger,M. 14092. (2020) An advanced human intestinal coculture model reveals 24. de Groot,N.S. and Torrent Burgas,M. (2019) ACoordinated Response compartmentalized host and pathogen strategies during Salmonella at The Transcriptome and Interactome Level is Required to Ensure infection. mBio, 11, e03348. Uropathogenic. Microorganisms, 7, 292. 11. Damron,F.H., Oglesby-Sherrouse,A.G., Wilks,A. and Barbier,M. 25. Rendón,J.M., Lang,B., Tartaglia,G.G. and Burgas,M.T. (2020) (2016) Dual-seq transcriptomics reveals the battle for iron during BacFITBase: a database to assess the relevance of bacterial genes Pseudomonas aeruginosa acute murine pneumonia. Sci. Rep., 6, during host infection. Nucleic Acids Res., 48, D511–D516. 39172. 26. Wolf,T., Kammer,P., Brunke,S. and Linde,J. (2018) Two’s company: 12. Griesenauer,B., Tran,T.M., Fortney,K.R., Janowicz,D.M., studying interspecies relationships with dual RNA-seq. Curr. Opin. Johnson,P., Gao,H., Barnes,S., Wilson,L.S., Liu,Y. and Spinola,S.M. Microbiol., 42, 7–12. (2019) Determination of an interaction network between an extracellular bacterial pathogen and the human host. mBio, 10, e01193.