Implementation of the Generalized Parallel Replica algorithm

I am working on the Parallel Replica (ParRep) algorithm, first introduced by Arthur Voter, further extended by Lelièvre et al. (, which relies on the convergence of stochastic processes (e.g. Langevin Dynamics) to a Quasi Stationary Distribution (QSD). My task is to develop a robust, flexible and fast implementation that one can apply to any chemical or biological system.

Diagram of an MPI based parRep implementation


Ideal (red, geometric law) and parRep estimated (histogram) distribution of the escape time from a metastable domain

In November 2017, I have obtained 800,000 CPU-hours on the national OCCIGEN supercomputer, administrated by the CINES : I will use this time for investigating protein-ligand systems using our parRep implementation.

The FKBP-DMSO protein-ligand system, a good candidate for validating the parRep algorithm : we want an accurate estimate of the mean dissociation time.