Annotating activation/inhibition relationships to protein-protein interactions using gene ontology

Abstract

Cells react to stimuli through signaling pathways, in which proteins physically bind to each other to propagate signals. To reconstruct signaling pathways from Protein-Protein Interactions (PPIs), activation/inhibition relationships should be annotated to PPIs. In this research, we annotated activation/inhibition relationships to PPIs by using regulates, positively regulates, and negatively regulates relations in Gene Ontology (GO). To annotate activation/inhibition relationships, we firstly predicted directions of undirected PPIs, and then predicted sign for each directed PPI. For predicting directions, we trained two logistic regression models for predicting whether signal can flows in left-to-right direction and right-to-left direction, respectively. For predicting signs, we trained two logistic regression models for predicting whether a directed PPI can act as activation and inhibition, respectively. Independent test results show that the proposed method outperforms previous GO-based method for predicting activation/inhibition relationships of PPIs. The activation/inhibition relationships of PPIs enrich PPI network, enhancing our understanding of signaling pathways and providing it as a valuable resource in systems biology.