An evaluvation of discrete and continuum search techniques for conformational analysis of side chains in proteins

Abstract Methodology for calculation of side‐chain conformations in proteins is evaluated. The role and impact of corrections to idealized rotameric structures are considered, by incorporating methods for torsional optimization into rotamer‐packing algorithms. Off‐rotamer corrections given by continuum torsional optimization improve, over simpler rotamer‐packing procedures, the accuracy with which the conformations of side chains of buried amino acids can be predicted. The analogy between protein side‐chain calculations and spin systems is explored by adapting spin simulation methods to side‐chain packing algorithms. Implementations of mean‐field and heat‐bath algorithms for side‐chain packing are described and their performance tested. The procedures introduced here address the combinatorial problem in an efficient and reasonably effective manner, as evidenced by analysis of their convergence properties. Application of refined protocols yields overall prediction accuracies of 80% for χ 1 and 68percnt; for χ 1,2 pairs for a test set of 60 proteins, using a 40° cutoff to define correct placement. For buried amino acids (defined as having less than 30% relative solvent accessibility) the prediction accuracies increase to 88percnt; for χ 1 and 79percnt; for χ 1,2 pairs. The influence of the form of the potential energy function is studied by comparing results obtained with 12‐6 and 9‐6 potentials. The 9‐6 form leads to more accurate results. Detailed comparison with previous work is presented, and the effect of combinatorial packing steps is shown to be important for all but the smallest proteins. © 1995 John Wiley & Sons, Inc.