Protein folding using contact maps

TL;DR

This paper explores a novel approach to protein folding using contact maps and Monte Carlo dynamics, but finds that simple free energy approximations are inadequate for accurate predictions.

Contribution

Introduces a contact map-based computational framework for protein folding, including methods for conformation reconstruction and energy parameter derivation.

Findings

Simple free energy approximations are unsuitable for protein folding.

Demonstrated energy minimization on crambin and immunoglobulins.

Developed Monte Carlo dynamics in contact map space.

Abstract

We present the development of the idea to use dynamics in the space of contact maps as a computational approach to the protein folding problem. We first introduce two important technical ingredients, the reconstruction of a three dimensional conformation from a contact map and the Monte Carlo dynamics in contact map space. We then discuss two approximations to the free energy of the contact maps and a method to derive energy parameters based on perceptron learning. Finally we present results, first for predictions based on threading and then for energy minimization of crambin and of a set of 6 immunoglobulins. The main result is that we proved that the two simple approximations we studied for the free energy are not suitable for protein folding. Perspectives are discussed in the last section.