Optimizing the performance of bias-exchange metadynamics: folding a 48-residue LysM domain using a coarse-grained model

Computer simulation of conformational transitions in biomolecules, such as protein folding, is considered one of the main goals of computational chemistry. Due to the complexity of the systems, a quantitative description can only be provided at the price of a significant computational cost. A powerful methodology, called bias exchange metadynamics (BE-META) [16], has been recently developed. The approach combines replica exchange [15] with metadynamics [14], and allows exploring the free energy landscape of complex systems, like biomolecules. The primary objective of this thesis is to improve further this promising technique. This will be accomplished by searching for the optimal set of parameters (e.g. collective variables and exchange time) that enable the folding of a small protein 1E0G (48 amino acids) in the shortest possible time, using a coarse-grain force field (UNRES [18]). It will be shown that BE-META allows the accurate recon- struction of the folding free energies of 1E0G, with a small computational effort in comparison with other techniques like MREMD, and that a suffcient number of collective variables are necessary to increase the capability of each replica to diffuse through the conformational space.