Ensemble based convergence assessment of biomolecular trajectories

Abstract

Assessing the convergence of a biomolecular simulation is an essential part of any computational investigation. This is because many important quantities (e.g., free energy differences) depend on the relative populations of different conformers; insufficient convergence translates into systematic errors. Here we present a simple method to self-consistently assess the convergence of a simulation. Standard clustering methods first generate a set of reference structures to any desired precision. The trajectory is then classified by proximity to the reference structures, yielding a one-dimensional histogram of structurally distinct populations. Comparing ensembles of different trajectories (or different parts of the same trajectory) built with the same reference structures provides a sensitive, quantitative measure of convergence. Please note: this is a preliminary manuscript, and should be read as such. Comments are most welcome, especially regarding pertinent prior work.

Preliminary manuscript, comments desired and welcome

References in corpus (1)

• Parallel Tempering Algorithm for Conformational Studies of Biological Molecules

Cited by in corpus (4)

• The structural de-correlation time: A robust statistical measure of convergence of biomolecular simulations
• Efficient equilibrium sampling of all-atom peptides using library-based Monte Carlo
• Extending fragment-based free energy calculations with library Monte Carlo simulation: Annealing in interaction space
• A Second Look at Canonical Sampling of Biomolecules using Replica Exchange Simulation