Monte Carlo simulations of protein folding using inexact potentials: how accurate must parameters be in order to preserve the essential features of the energy landscape?

Abstract

Monte Carlo simulations of the cubic lattice protein model with engineered sequences were performed in order to address the issue of potential accuracy required for folding. The potential used for sequence selection played the role of the 'real' potential and different levels of inaccuracy were introduced by addition of noise. The dependence of successful folding probability on potential noise was found to be sigmoidal and sequence-specific and can be described by an expression analytically derived from a simple theoretical model in which the density of states of the system contains a continuous region approximated by a Gaussian distribution separated from the unique native conformation by a large energy gap. The decrease in folding probability with potential inaccuracy results from an average decrease in the energy gap. Sequences with large energy gaps support larger inaccuracies while retaining the ability to fold properly. As the energy gap is known to correlate with thermal stability, we suggest a simple criterion for specific real sequence selection in order to maximize success probability in realistic folding simulations.