A literature-based approach for curating gene signatures in multifaceted diseases

摘要

Workflow of constructing a gene panel that putatively contains subtype signatures of Inflammatory Bowel Disease (IBD). a network of the predicted protein–protein interactions (STRING) inferred from the genes associated with immune responses and IBD highlights the complexity and difficulty of making logical interpretation (network image in top-left corner). In order to simplify the process, we devised six gene lists from different sources. Details on the methods used to retrieve the genes, the number of genes in the lists, and the terms used onwards are shown (Gene Lists). Then, we submitted the lists to Venn analysis which resulted in 21 intersections (Venn Analysis). The common (i.e. IBD core genes) and the unique lists for CD, UC, and IBDU were submitted to Literature Lab to obtain pathways and diseases association scores; each gene is ranked based on its contribution weight to a score (Extract Unique Groups–Literature Lab). In the network analysis, the top-ranking genes (those with > 5% contribution to the association) informed which nodes to expand to the primary and secondary nodes. We used both GeneMania and STRING (shown here) to obtain those networks. In the network shown, the edges depict the known protein interactions based on knowledge from curated databases (blue edges), experimentally determined (pink edges), and co-expression data (black edges). Genes without shared pathways are shown as independent nodes. The amalgamation of the gene selections from all the common and subtypes-specific genes amounts to 142 putative target genes