Building protein lattice models using self-consistent mean field theory

Abstract

An optimization protocol for modeling protein structures on lattice is proposed which is based on self-consistent mean field (SCMF) theory. In this procedure, the protein residues are supposed to be independent, and their possible positions are given by a list of lattice sites. To do this, an effective larger system is considered, in which each residue i is supposed to occupy all possible sites j, each with a weight V(i,j) stored in the so-called lattice probability matrix V. The effective energy of the system is computed, and iteratively minimized with respect to the weights V, the lattice sites being fixed in space. The final self-consistent V matrix describes the conformational space available to the protein, based on the energy function implemented. This energy function contains two types of terms, namely simple geometric terms which ensure bond connectivity and prevent chain intersection, and energy terms specific to the problem of interest. The application of the above protocol to building a lattice model of a protein, given its three dimensional structure, is discussed and compared with other lattice fitting procedures.