Conformational classification of short backbone fragments in globular proteins and its use for coding backbone conformations

Abstract

An objective and systematic method of coding backbone conformations of proteins is developed. For this purpose polypeptide backbone is regarded to consist of two types of overlapping structural units, viz. a (φ, ψ) fragment (C i αC′ONC i+1 αC′ONC i+2 α), and a (ψ, φ) fragment (NC i αC′ONC i+1 αC′). By means of the principal component analysis, these (φ, ψ) and (ψ, φ) fragments are found to form five and six distinct clusters in respective high-dimensional conformational space with only a small number of exceptional unclassified fragments. Boundaries of clusters are defined by multi-dimensional ellipsoids. This classification of conformations of short backbone fragments is used to code protein backbone three-dimensional structures. In particular, conformations of peptide fragments consisting of four or six residues are coded and analyzed in detail. This structural code allows us to distinguish such features of protein backbone structures that are not retained in the usual secondary structural representation based on patterns of backbone hydrogen bonds, e.g., various types of four-residue turns. Similarity index, based on the number of different one-letter codes in a pair of structural codes, is a powerful measure of conformational similarity. When combined with another similarity measure, atomic root-mean-square distance, accurate similarity between a pair of conformations of fragments can be detected.