Uncertainty quantification in classical molecular dynamics

Abstract

Molecular dynamics simulation is now a widespread approach for understanding complex systems on the atomistic scale. It finds applications from physics and chemistry to engineering, life and medical science. In the last decade, the approach has begun to advance from being a computer-based means of rationalizing experimental observations to producing apparently credible predictions for a number of real-world applications within industrial sectors such as advanced materials and drug discovery. However, key aspects concerning the reproducibility of the method have not kept pace with the speed of its uptake in the scientific community. Here, we present a discussion of uncertainty quantification for molecular dynamics simulation designed to endow the method with better error estimates that will enable it to be used to report actionable results. The approach adopted is a standard one in the field of uncertainty quantification, namely using ensemble methods, in which a sufficiently large number of replicas are run concurrently, from which reliable statistics can be extracted. Indeed, because molecular dynamics is intrinsically chaotic, the need to use ensemble methods is fundamental and holds regardless of the duration of the simulations performed. We discuss the approach and illustrate it in a range of applications from materials science to ligand–protein binding free energy estimation. This article is part of the theme issue ‘Reliability and reproducibility in computational science: implementing verification, validation and uncertainty quantification in silico ’.

Hit-to-lead and lead optimization binding free energy calculations for G protein-coupled receptors

“…One of these, ESMACS [2,14,16,17,19] (enhanced sampling of molecular dynamics with approximation of continuum solvent), uses ensembles of multiple and typically relatively short duration simulations to calculate absolute binding free energies with high precision. There is a wealth of evidence in the literature and in unpublished work that, under equilibrium conditions, multiple short MD simulations sample better than a single long MD simulation and provide a meaningful uncertainty of the results [8,9,20–22]. Under general non-equilibrium conditions, ensembles are essential since there is then no meaning to time averaging.…”

“…It should be noted that single long time simulations are inaccurate, as we have explained before (e.g. [22]). Ensembles are required for all MD simulations [20,22], as precise predictions, along with their uncertainties, can be obtained only when the most relevant conformations have been extensively sampled.…”

“…[22]). Ensembles are required for all MD simulations [20,22], as precise predictions, along with their uncertainties, can be obtained only when the most relevant conformations have been extensively sampled. A particular benefit of ESMACS is the freedom to choose multiple trajectory versions to enhance predictions and provide qualitative information about and insight into the associated binding mechanisms.…”

The effect of protein mutations on drug binding suggests ensuing personalised drug selection

“…We used ensemble-based ESMACS and TIES for the free energy calculations. Extensive studies have confirmed that the most effective and reliable computational route to reproducible predictions using MD simulation can be achieved using ensemble methods 23 , 39 – 41 . A set of independent MD simulations are employed to obtain the required averages and associated uncertainties.…”

Reliability and reproducibility in computational science: implementing validation, verification and uncertainty quantification in silico

“…Wan et al [5] discuss the quantification of uncertainty in simulations that are based on classical molecular dynamics. The paper addresses simulations in a wide range of applications from binding affinity calculations for drug discovery to properties prediction within condensed matter and materials.…”