Identifying Dominant Amino Acid Pairs of Known Protein-Protein Interactions via K-Means Clustering

Abstract

© 2017 IEEE. Biological functions in all living cells are performed by protein-protein interactions since they form cells and control function mechanisms. Thus, identifying pairs of protein-protein interactions would be very useful, but it is not an easy task. But, doing a wet lab consumes huge amount of resources whereas using computational methods is highly challenging since they may introduce high false positives. Since a protein is a sequence of amino acids, a protein interaction would be influenced by some interactions of amino acids, and the identification of outstanding interacting pairs would give insightful meaning into how a pair of proteins interacts. This paper proposes a novel method to analyze a set of well-known protein-protein interactions for identifying a set of strong amino acid pairs that may influence the interaction. We calculate amino acid correlation values via Pearson's correlation, and use K-means clustering to group a set of outstanding amino acid pairs based on correlation values. The experimental results for 10 sets of protein interaction networks can identify a number of strong amino acid pairs among them.