Rheumatoid_Arthritis_EXTENDED_INTERACTOME

Curated collection of molecules from high-throughput screens of diverse (multi-omic) biochemical origin, experimentally associated to RA. Starting from such collection the RA-related protein-protein interaction (PPI) networks (interactomes)based on experimental PPI data have been generated.
The datasets used to construct the map are gathered from 13 different sources from databases and literature (Table 1). We included molecules experimentally associated to RA from manual curation of literature sources (core dataset, CD, 377 proteins), and additional molecules and pathways strongly yet not explicitly associated to RA (extended dataset, ED, 4709 proteins).
The extended dataset (ED, that includes CD) is composed of 4709 proteins, which are involved in a broader sense in the onset and development of RA, such as proteins participating in signaling pathways or cascades of recognized importance for RA. This extension provides a more general setting for the molecular framing of RA, and offers a larger network to operate on, with more relevant statistics and analyses, giving account for contributions coming from entities that may have been neglected or that are not experimentally related to RA, but that participate to the inception of the disease. In addition to the proteins of the coredataset, we added eight main subsets, as follows (Data Sheet 1, Tables S3A–E, S7–S13):
3A-B-C-D-E) in retrieving data from Mcinnes and Schett (2011)and references cited there, we considered that some of the key proteins can be “hidden” inside the signaling pathways involved in the disease. In order to take into account such potentially important and usually neglected elements, we expanded subset 3 of CD by a pathway enrichment analysis process, using the genes listed in Mcinnes and Schett (2011)Tables 1, 2. To populate these five subsets, the selected genes have been input in the pathway over-representation analysis(ORA) tool of InnateDB, one of the most comprehensive sources of pathways data available (Lynn et al., 2008; Breuer et al., 2013). Pathway ORA has been performed on InnateDB using hypergeometric distribution for p-value computation and Benjamini–Hochberg correction method for multiple hypothesis testing. All the proteins participating to such over-represented pathways were then included. We retrieved respectively: 39 enriched pathways accounting for 1248 proteins (subset 3A), 14 pathways and 283 proteins (3B), 46 pathways and 1536 proteins (3C), 5 pathways and 472 proteins (3D), and 92 pathways and 1837 proteins (3E), all collected in Data Sheet 1, Tables S3A–E;
7) Genes derived from the transcriptional RA map in Wu et al. (2010)(Data Sheet 1, Table S7);
8) RA-related miRNA-regulated genes: experimentally validated target genes of all miRNAs that are associated to RA in the database miRWalk (Dweep et al., 2011) (search mode: holistic view of validated disease-miRNA interactions; web reference: http://www.umm.uni-heidelberg.de/apps/zmf/mirwalk/disease.html; query keywords: ArthritisAND Rheumatic diseases) (Data Sheet 1, Table S8);
9A,B) gene expression profiles of RA patients and healthy controls were searched on Gene Expression Omnibus (GEO, (Barrett et al., 2011) http://www.ncbi.nlm.nih.gov/geo/) with the query [“rheumatoid arthritis” AND “(synovi* OR blood)”] (i.e., in synovial tissue and/or blood). In order to include only highly consistent information, datasets without pre-treatment samples, with no details about the therapy and no raw data were filtered out. Human PBMCs collected and processed by Affymetrix technology were selected, leaving only one dataset out of the initial 61, GSE7524, which contains transcriptomic profiles of 2 healthy controls, 2 before and 2 after anti-TNFα treatment samples. Affymetrix Human Genome U133A Array was used to measure the expression levels of ~14,500 well-characterized human genes. The raw data were pre-processed using affypackage (Gautier et al., 2004) in R (http://www.r-project.org/), normalized using robust multi-array average (rma) (Irizarry et al., 2003) and for multiple probes corresponding to the same gene, the probe with the highest standard variation across all samples was used to represent the gene. Differentially expressed genes [fold-change (Murie et al., 2009) =2] were identified with the comparison between the 2 healthy controls and the 2 before anti-TNFα treatment samples resulting in 646 genes differentially expressed, among which 440 genes (451 proteins) were down-regulated and 206 genes (210 proteins) were up-regulated (Data Sheet 1, Tables S9A,B);
10) Proteins related to the inflammasome, a multiprotein oligomer responsible for activation of inflammatory processes proteins, which is also known to be activated from the bacterium P. Gingivalis, among others, and recognized to play a relevant role in RA (Sidiropoulos et al., 2008; Kolly et al., 2010; Farquharson et al., 2012; Mathews et al., 2013) (Data Sheet 1, Table S10). This set was retrieved using ORAas described in 3A-B-C-D-E;
11) Adenosine receptors and related proteins, known to be involved in RA (Varani et al., 2010, 2011; Vincenzi et al., 2013) and possibly at the basis of the mechanism of action of methotrexate, first-line therapy for the treatment of RA (Stamp et al., 2012) (Data Sheet 1, Table S11). This set was retrieved using ORAas in 3A-B-C-D-E and 10;
12) The large family of G Protein Coupled Receptors (GPCRs) (Hutchings et al., 2010; Lozupone et al., 2012; Maynard et al., 2012; Tremaroli and Backhed, 2012), pertaining to host-microbiome interface proteins (grouped in a separate set from 13 due to their numerosity), retrieved from http://www.iuphar-db.org/DATABASE/ReceptorFamiliesForward?type=GPCR(Sharman et al., 2013) (Data Sheet 1, Table S12);
13) The set of host-microbiome interacting proteins, manually curated from recent reviews (Lozupone et al., 2012; Maynard et al., 2012; Tremaroli and Backhed, 2012), to describe the bridge between innate immunity (altered in RA) and the GI microbiome [known to be involved in immune diseases in general and in RA in particular (Scher and Abramson, 2011)]. Globally this dataset accounts for the Toll-like Receptor family (TLRs), the mucin proteins family, selected Immunoglobulins (Ig) and their receptors, among others (Data Sheet 1, Table S13).
Datasets are integrated at the PPI level as peers to avoid introducing any bias a prioriin the network construction and to warrant that these data are connected in a biologically meaningful way. Protein-protein interactions were retrieved in Cytoscape from the Agile Protein Interaction DataAnalyzer database (APID, Prieto and De Las Rivas, 2006) that includes all known experimentally validated protein-protein interactions from BIND, BioGRID, DIP, HPRD, IntAct and MINT databases, accessed via the APID2NET (Hernandez-Toro et al., 2007) plugin. This process lead to the definitions of, respectively, the core interactome (CI, 303 proteins, 597 interactions, high resolution Image S1) and the extended interactome (EI, 3783 proteins, 24457 interactions, high resolution Image S2). Discussion on caveats and choices of original sources can be found in Tieri and Nardini (2013)