Weighted Ensemble is a flexible framework for enhanced sampling methods, where a set of trajectories is run in parallel, and trajectories are periodically cloned (to increase sampling in low-probability regions) or merged (to decrease sampling of high-probability regions). The ensemble of trajectories are each given a weight (wi), which are halved upon cloning, and are summed upon merging, locally conserving probability. This was originally implemented with simple binning schemes, which are not ideal for sampling high-dimensional free energy surfaces. Our lab has developed “cloning and merging” frameworks (also called “resampling”) that do not rely on a single collective variable or order parameter, but instead use a measure of distance between members of the trajectory set, which could be defined in a high-dimensional space. To facilitate the broad adoption of these methods our lab is introducing “wepy” (weighted ensemble in python). This is a modular framework for running weighted ensemble simulations using the high-dimensional resampling algorithms (e.g. WExplore, or REVO), written in pure python. Wepy comes pre-loaded with a set of analysis tools that are particularly useful for weighted ensemble data. For instance, repeated cloning and merging operations result in a branching trajectory tree that grows upwards as the simulation progresses. Wepy stores the trajectory and resampling data in an HDF5 format that is cognizant of the tree structure, allowing for quick querying of simulation properties, and construction of time-lagged Markov transition matrices that take into account trajectory branching. Here we give a broad overview of the wepy software package, show example use cases, and summarize recent results obtained with the wepy package using newly added distance metrics and resampling algorithms. Together with a modest size GPU cluster, wepy can allow access to extremely long timescale events, with mean first passage times as long as of tens of minutes! We are excited to help new users adopt this software to explore free energy landscapes for their systems of interest.

Wepy: Tool for exploring rough free energy landscapes

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.