A Fully Automatic and Efficient Methodology for Peptide Activity Identification Using Their 3D Conformations

Over the past decades, understanding of peptides and proteins biological functions has been an active research topic. Latest works in this field have suggested that protein conformations may be a key feature for gaining insights into functions. However, analyzing small highly flexible chunks, namely oligopeptides made handful amino acids, remains challenging because their dynamics wide range conformations. In paper, statistical methodology based on unsupervised learning is proposed 3D elastin-derived peptides. The goal study twofold: first, it aimed at identifying most frequent each peptide to stability. Second, important, comparing main different identify “signature” than can linked activity. strength present work propose method confirmation recognition not affected by rotations or translations and, hence, avoids use complex superposition methods. addition, approach uses Kernel PCA eliminate atypical Due instability those peptides, removing outliers crucial since they dramatically impact clustering results. To extract conformations, we hierarchical method. Eventually, activity detector defined comparison conformation found interests are fully automatic method, second, does require any additional information expertise third, accurately make enabling given Experimental results large dataset highlight relevance efficiency