Beyond second-long trajectory of the Trp-cage peptide generated using a Kinetic Monte Carlo model derived from molecular dynamics
TL;DR
A kinetic Monte Carlo modeling approach to describe the stochastic dynamics of a peptide molecule spanning nanosecond to second timescales is presented, and it is shown that with the correlations built-in, the KMC model closely matches MD.
Abstract
We present a kinetic Monte Carlo (KMC) modeling approach to describe the stochastic dynamics of a peptide molecule spanning nanosecond to second timescales. The dynamics of protein conformational changes is interpreted at a local level in terms of dihedral transitions. Taking Trp-cage miniprotein as an example, the KMC model"learns"about the transitions from multiple MD trajectories. Training is based on local divide-and-conquer strategy that identifies the discretized backbone dihedral states as building blocks for the conformational space, along with associated transition rates of dihedral flips to describe the conformational state-to-state dynamics. A key feature in our approach is the incorporation of backbone correlations, such that rates are conditioned on the local environment and steric coupling. We show that with the correlations built-in, the KMC model closely matches MD. Such an approach is shown to reach second timescales in a few CPU hours on a standard desktop computer, and can easily yield multiple stochastic realizations of the conformational dynamics. Our KMC model construction scheme should be generally applicable to a wide range of proteins, and can be used for bridging local flexibility to protein-wide dynamics.