Progressing from folding trajectories to transition state ensemble in proteins

A general procedure for determining the nature of the transition state ensemble (TSE) in proteins is used to probe the heterogeneity of TSE in two models of alpha/beta proteins. The sequences, which differ only in the chain connectivity, are modeled using alpha-carbon representation. The structures in the TSE are determined using a progress variable clustering (PVC) algorithm that does not depend on the choice of the reaction coordinate. Folding simulations starting from these transition state structures partition with nearly equal probability into the native state or unfolded state. Our fully automated PVC algorithm, which utilizes only the time series in the folding trajectories, is also computationally efficient. The results for the model alpha/beta proteins show that the TSE can be divided into a small number of clusters (two or three) consisting of structurally related conformations. For both sequences the conformations in the dominant transition cluster are compact and native-like with the secondary structure elements fully formed. Although none of tertiary contacts is completely formed in the transition region, few occur with high probability. This suggests that there are multiple (more than one) nuclei that connect the unfolded and native states. Overstabilizing the high probability nucleation contacts retards folding rates. The results for both the models also suggest that, if the transition region is close to the native state, then permutation of the secondary structural elements does not alter the nature of the transition state. (C) 2004 Elsevier B.V. All rights reserved.