An effective machine learning-based model for the prediction of protein–protein interaction sites in health systems

Protein is a vital biomolecule that accomplishes distinct biological activities by interacting with other proteins in complex biological systems. The protein–protein interaction (PPI) sites hot spot characterization holds preliminary importance in drug discovery as well as in the comprehension of the cellular signaling phenomenon. Looking at the significance of PPIs, an intelligent prediction system based on the notion of fuzzy logic “PPIs-FuzzyKNN” is developed for PPI sites identification. Here, protein sequences are transformed into an equal length of numerical descriptors by using physicochemical properties of amino acids and a position-specific scoring matrix. Here, we have utilized conventional machine learning algorithms as well as fuzzy k-nearest neighbors. The results of the model are assessed via a tenfold cross-validation test. The proposed model PPIs-FuzzyKNN obtained 91.20, 92.65, and 93.50% of accuracy on the three different datasets, namely Dtestset72, PDBtestset164, and Dset186, respectively. The results exhibited that the outcomes of the proposed model are outstanding and persistent in all datasets, so far, compared to the literature. Consequently, it will not only play a leading role in the accurate identification of PPI sites but also becomes a rudimentary tool for the research community.