UNDERSTANDING THE RECOGNITION OF PROTEIN STRUCTURAL CLASSES BY AMINO-ACID-COMPOSITION

(K)nowledge of amino acid composition, alone, is verified here to be s ufficient for recognizing the structural class, alpha, beta, alpha + b eta, or alpha/beta of a given protein with an accuracy of 81%. This is supported by results from exhaustive enumerations of all conformation s for all sequences of simple, compact lattice models consisting of tw o types (hydrophobic and polar) of residues. Different compositions ex hibit strong affinities for certain folds. Within the limits of validi ty of the lattice models, two factors appear to determine the choice o f particular folds: 1) the coordination numbers of individual sites an d 2) the size and geometry of non-bonded clusters. These two propertie s, collectively termed the distribution of nonbonded contacts, are qua ntitatively assessed by an eigenvalue analysis of the so-called Kirchh off or adjacency matrices obtained by considering the non-bonded inter actions on a lattice. The analysis permits the identification of confo rmations that possess the same distribution of non-bonded contacts. Fu rthermore, some distributions of non-bonded contacts are favored entro pically, due to their high degeneracies. Thus, a competition between e nthalpic and entropic effects is effective in determining the choice o f a distribution for a given composition, Based on these findings, an analysis of non-bonded contacts in protein structures was made. The an alysis shows that proteins belonging to the four distinct folding clas ses exhibit significant differences in their distributions of non-bond ed contacts, which more directly explains the success in predicting st ructural class from amino acid composition. (C) 1997 Wiley-Liss, Inc.d agger