Figure 1. Workflow of the study and the comparisons with other data.

(a) Workflow of protein-protein interaction prediction by Naïve Bayes integration and PPI network analyses. PPIs were predicted with seven different features: smallest shared biological process (SSBP), microarray gene expression profiles, domain interaction, phylogenetic profile, gene fusion, gene neighborhood and gene cluster. Using gold standard positive dataset (GSP) and gold standard negative dataset (GSN) sets as positive and negative data, we calculated the likelihood ratio (LR) for each PPI in each feature, and then integrated these LRs with Naïve Bayes approach. (b) The overlapped and unique PPIs predicted in this study and two other similar studies, InteroPORC and SynechoNET. (c) Evaluation and comparison of our prediction, InteroPORC and SynechoNET using an independent data, STRING scores.