Microscopic interpretation of folding φ-values using the transition path ensemble

All-atom molecular dynamics simulations now allow us to create movies of proteins folding and unfolding. However, it is difficult to assess the accuracy of the folding mechanisms observed because experiments cannot yet directly resolve events occurring along the transition paths between unfolded and folded states. Protein folding.-values provide residue-resolved information about folding mechanisms by comparing the effects of mutations on folding rates and stability, but determining.-values by separately simulating mutant proteins would be computationally demanding and prone to large statistical errors. Here we use transition path theory to develop a method for computing phi-values directly from the transition path ensemble, without the need for additional simulations. This path-based approach uses the full transition path information available from equilibrium folding and unfolding trajectories, or from transition path sampling, and does not require identification of folding transition states. Applying our approach to a set of simulations of 10 small proteins by Shaw and coworkers [Lindorff-Larsen K, Piana S, Dror RO, Shaw DE (2011) Science 334(6055): 517-520; Piana S, Lindorff-Larsen K, Shaw DE (2011) Biophys J 100(9): L47-L49; and Piana S, Lindorff-Larsen K, Shaw DE (2013) Proc Natl Acad Sci USA 110(15): 5915-5920], we find good agreement with experiments in most cases where data are available. We can further resolve the contributions to fractional.-values coming from partial contact formation versus transition path heterogeneity. Although in some cases, there is substantial heterogeneity of folding mechanism, in others, such as Ubiquitin, the mechanism is strongly conserved.