Time-averaged predictions of folded and misfolded peptides using a reduced physicochemical model

Energy-based methods for calculating time-averaged peptide structures are important for rational peptide design, for defining local structure propensities in large protein chains, and for exploring the sequence determinants of amyloid formation. High-end methods are currently too slow to be practicable, and will remain so for the foreseeable future. The challenge is to create a method that runs quickly on limited computer resources and emulates reality sufficiently well. We have developed a simplified off-lattice protein model, incorporating semi-empirical physicochemical potentials, and combined it with an efficient Monte Carlo method for calculating time-averaged peptide structures. Reasonably accurate predictions are found for a set of small alpha-helical and beta-hairpin peptides, and we demonstrate a potential application in measuring local structure propensities in protein chains. Time-averaged structures have also been calculated for a set of small peptides known to form beta-amyloid fibrils. The simulations were of three interacting peptides, and in each case the time-averaged structure describes a three-stranded beta-sheet. The performance of our method in measuring the propensities of small peptides to self-associate into possible prefibrillar species compares favorably with more detailed and CPU-intensive approaches.