Leveraging a Structure-Based and Learning-Based Predictor Using Various Feature Groups in Bioinformatics (Case Study: Protein-Peptide Region Residue-Level Interaction)

Predicting protein-peptide interactions is crucial for understanding cellular activities, researching protein functions, designing new drugs, studying abnormal cell behavior, and addressing human diseases. Conventional experimental techniques for identifying the residues in protein-peptide interaction regions are labor-intensive and costly. Therefore, determining these binding region residues using computational methods would serve as a valuable and complementary tool. Our research introduces RSPPRI, a computational method that combines a Residual Neural Network and a Support Vector Machine to accurately identify the residues in the binding regions of protein-peptide interactions. To achieve this, we extract various feature groups from protein structures, including evolutionary, physicochemical, and structure-based features.