Analysis of Multiple Protein Alignments Using 3D-Structural Information on the Orientation of Amino Acid Side-Chains

Abstract

Multiple alignment of amino acid sequences of homologous proteins is a key tool in state-of-the-art bioinformatics and evolutionary analysis. Differences in the spatial orientation of amino acid side-chains can predetermine significant functional diversity among members of one superfamily; however, this is usually not taken into account in any way when constructing alignments and during subsequent comparative analysis. First of all, this is due to the limitation of existing algorithms, which are guided by the biochemical similarity of the "alphabet" of amino acid substitutions and either do not use information about the 3D-structural organization of proteins at all, or are limited to comparing the backbone only (i.e., the atoms of the main-chain). In this work, for the first time, we introduce new software for a systematic analysis of specific orientations of amino acid side-chains in equivalent positions of homologous protein structures. The program is intended to assist the analysis of protein multiple sequence alignments. The new algorithm, based on the machine learning HDBSCAN method, can identify statistically significant differences in the side-chain orientations and classify them into subfamilies at each position of multiple alignment. The method has been tested on a wide set of real biological data. The results allow us to speak of the specific orientation of amino acid side-chains as a common phenomenon that requires further study and deserves attention in a comparative analysis of functionally diverse protein superfamilies. The software is freely available at https://github.com/LimoninaDaria/Sub-family-Specific-Sidechain-Orientations.