Protein tertiary structure prediction via global optimization of reduced model potential energy functions

We have implemented several methods for protein tertiary structure prediction to low resolution. Structure is predicted by globally minimizing knowledge-based potential energy functions that employ simplified geometric models of the polypeptide chain. The approximate location and configuration of secondary structural units are assumed to be known and enter the calculations through conformational restraints. Performance of the approaches based on the Monte Carlo plus Minimization and Branch and Bound algorithms is assessed by comparing results for a large and structurally diverse test set. Inclusion of explicit protein size-dependent potential energy function terms results in a qualitative improvement of the performance, rendering the methods applicable to alpha-helical and mixed-alpha/beta proteins up to 100 residues in size. Several examples illustrate the use of the methods in the context of ab initio tertiary structure prediction, where secondary structure is predicted with existing algorithms.