A Machine Learning Method to Characterize Conformational Changes of Amino Acids in Proteins

Abstract

AbstractAmino acid dynamics are significant in determining the overall function, structure, stability, and activity of proteins. However, atomic-level descriptions of the structural features of proteins are limited by the current resolutions of experimental and computational techniques. In this study, we developed a machine learning (ML) framework for characterizing the individual aminoacids dynamic in a protein and compute its contribution to the overall function of proteins. This framewor identifies specific types of angular features in amino acids, such as bimodal-switch residues. It can assist in the analysis of various protein characteristics and provide valuable insights into the dynamic behavior of individual amino acids within a protein structure. We found that there is a strong correlation between a specific type of bimodal-switch residues and the global features in proteins. This knowledge can help us to identify key residues that are strongly correlated to the overall function of the protein.