Assessment of semiempirical enthalpy of formation in solution as an effective energy function to discriminate native‐like structures in protein decoy sets

In this work, we tested the PM6, PM6‐DH+, PM6‐D3, and PM7 enthalpies of formation in aqueous solution as scoring functions across 33 decoy sets to discriminate native structures or good models in a decoy set. In each set these semiempirical quantum chemistry methods were compared according to enthalpic and geometric criteria. Enthalpically, we compared the methods according to how much lower was the enthalpy of each native, when compared with the mean enthalpy of its set. Geometrically, we compared the methods according to the fraction of native contacts ( Q ), which is a measure of geometric closeness between an arbitrary structure and the native. For each set and method, the Q of the best decoy was compared with the Q 0 , which is the Q of the decoy closest to the native in the set. It was shown that the PM7 method is able to assign larger energy differences between the native structure and the decoys in a set, arguably because of a better description of dispersion interactions, however PM6‐DH+ was slightly better than the rest at selecting geometrically good models in the absence of a native structure in the set. © 2016 The Authors. Journal of Computational Chemistry Published by Wiley Periodicals, Inc.