Determining protein–protein functional associations by functional rules based on gene ontology and KEGG pathway

• Published in: Biochimica et biophysica acta. Proteins and proteomics 2021-06, Vol.1869 (6), p.140621-140621, Article 140621
• Main Authors: Zhang, Yu-Hang, Zeng, Tao, Chen, Lei, Huang, Tao, Cai, Yu-Dong
• Format: Article
• Language: English
• Subjects: Decision tree; Gene ontology; KEGG pathway; Protein–protein functional association; Rule
• ISSN: 1570-9639; 1878-1454
• DOI: 10.1016/j.bbapap.2021.140621

Protein–protein interactions (PPIs) describe the direct physical contact of two proteins that usually results in specific biological functions or regulatory processes. The characterization and study of PPIs through the investigation of their pattern and principle have remained a question in biological studies. Various experimental and computational methods have been used for PPI studies, but most of them are based on the sequence similarity with current validated PPI participators or cellular localization patterns. Most methods ignore the fact that PPIs are defined by their specific biological functions. In this study, we constructed a novel rule-based computational method using gene ontology and KEGG pathway annotation of PPI participators that correspond to the complicated biological effects of PPIs. Our newly presented computational method identified a group of biological functions that are tightly associated with PPIs and provided a new function-based tool for PPI studies in a rule manner. •An explainable rule-based machine learning model of protein-protein interaction was built.•Each protein was represented by a binary GO and KEGG annotation vector.•We used the sum and the absolute difference of two protein vectors to represent a protein-protein pair.•The key GO and KEGG function features were identified using feature selection method.•The prediction rules of protein-protein interactions were learned using decision tree.