Figure 1: Construction and assessment of PseudomonasNet.

(A) A summary of the construction of an integrated co-functional network for P. aeruginosa. The co-functional links between P. aeruginosa genes were derived from nine diverse data sets: five P. aeruginosa co-functional networks, including co-citation (CC), co-expression (CX), correlation of protein domain profiles (DP), neighbourhood of bacterial orthologues (GN), and correlation of phylogenetic profiles (PG), and four associalog networks from co-citation and co-expression of E. coli orthologues and bacterial protein-protein interactions. PseudomonasNet was constructed based on a machine learning approach with reference gold-standard functional gene pairs that share Gene Ontology biological process annotations using a Bayesian data integration framework. (B) The integrated PseudomonasNet and individual component networks were assessed for precision via a comparison to KEGG pathway annotations. We measured the proportion of the gene pairs annotated by KEGG that share same pathway terms for every bin of 1,000 gene pairs from the highest score. The integrated network covers approximately 98% of the P. aeruginosa coding genes with superior precision to all individual component networks, which confirms the effectiveness of the data integration in the construction of the genome-scale functional network of the P. aeruginosa genes. (C) Network centralities of drug targets and essential genes are significantly higher than that of genomic average based on both degree- and betweenness-based scores in PseudomonasNet, which suggests that other genes with high centrality scores are good candidates for drug targets.