Program

I.10 -- Invited, IWTCP-3

Date: Wednesday, 29 July 2015

Time: 10h30 - 11h05

One-dimensional diffusion models for protein folding -- how good can they be?

Robert B. Best

National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA

Despite the vast number of degrees of freedom involved in protein folding, theoretical models have often assumed, with some success, that folding dynamics can be described as diffusion on a low-dimensional free energy surface. However, the validity of approximating dynamics as diffusion in a low-dimensional space could until recently only be indirectly assessed by comparing theory with experimental observables such as folding rates, or with highly simplified simulation models. The availability of all-atom atomistic folding simulations of ten small proteins, in explicit solvent, allows such assumptions to be directly tested. We find that for almost all of the proteins, it is possible to embed the folding dynamics rather accurately into diffusion on only a single progress variable, the fraction of native contacts (Q). Not only do the diffusion models, parametrized by position-dependent free energies F(Q) and diffusion coefficients D(Q), reproduce folding and unfolding rates, but also finer details of the dynamics, such as the transition-path durations, and propagators estimated directly from simulation. For the one exception, protein G, where the diffusion model for Q cannot capture the transition-path durations, we are able to find a modified coordinate, Qopt, by reweighting native contacts, which greatly improves the quality of the diffusion model. We find that the position-averaged diffusion coefficients on Q decrease with chain length, as anticipated from energy landscape theory. Overall, our results suggest that one-dimensional diffusion is a remarkably faithful model for protein folding dynamics, provided a suitable folding coordinate is chosen.

Presenter: Robert B. Best