How homologs can help to predict protein folds even though they cannot be predicted for individual sequences

At present, one cannot predict the 3D structure of a protein directly from its sequence alone mainly because of errors in the energy estimates. However, a recently developed simple analytical theory (Finkelstein, 1998) shows that using a set of homologs (i.e., chains with numerous amino acid mutations but with equal 3D folds) one can average the interaction energies over the homologs and predict their common 3D fold even when predictions for individual sequences are wrong because the energy parameters are known only approximately. In this work we verify this theoretical conclusion by computer simulations performed with simplified models of protein chains.