Multiobjective Approach to Gene Ontology-Based Protein-Protein Interaction Prediction

Protein-protein interactions (PPI) involve the binding of proteins to perform various biological functions, and accurately predicting them is vital for understanding protein behavior. Computational methods have gained popularity to reduce experimental costs, with studies exploring semantic similarity measures, particularly Gene Ontology (GO)-based ones. These measures assess the taxonomic similarity of GO terms to determine semantic similarity between protein pairs. Traditional taxonomic similarity measures are recognized tools for quantifying semantic similarity. This chapter introduces a Differential Evolution for Multiobjective Optimization (DEMO)-based technique to simultaneously optimize multiple criteria, including sensitivity, specificity, and the number of features. The study evaluates protein pair similarity measures, both traditional and graph-based, using GO-based semantic similarity measures for biological process, molecular function, and cellular component structures. A real-life dataset of yeast protein-protein interactions is employed to compare the performance of DEMO with existing approaches.