Unfolding Hidden Barriers by Active Enhanced Sampling

Abstract

Collective variable (CV) or order parameter based enhanced sampling algorithms have achieved great success due to their ability to efficiently explore the rough potential energy landscapes of complex systems. However, the degeneracy of microscopic configurations, originating from the orthogonal space perpendicular to the CVs, is likely to shadow "hidden barriers" and greatly reduce the efficiency of CV-based sampling. Here we demonstrate that systematic machine learning CV, through enhanced sampling, can iteratively lift such degeneracies on the fly. We introduce an active learning scheme that consists of a parametric CV learner based on deep neural network and a CV-based enhanced sampler. Our active enhanced sampling algorithm is capable of identifying the least informative regions based on a historical sample, forming a positive feedback loop between the CV learner and sampler. This approach is able to globally preserve kinetic characteristics by incrementally enhancing both sample completeness and CV quality.

Machine learning approaches for analyzing and enhancing molecular dynamics simulations

“…The The active enhanced sampling (AES) approach from Ref. [34] is similar to RAVE in the sense of iterating between enhanced sampling along a trial slow mode, and using the sampling to improve the slow mode definition. The enhanced sampling is carried out using well-tempered metadynamics [25].…”

Multiscale modeling: foundations, historical milestones, current status, and future prospects

“…The last decade has seen significant advances in the use of electronic structure calculations to train ML potentials for atomistic simulations capable of reaching large systems sizes and longtime scales with accurate and reliable energies and forces. More recently, ML approaches have proved useful in learning high-dimensional free energy surfaces [162,163], and in providing a low dimensional set of collective variables or CVs [164]. Some examples are discussed below.…”

Molecular Insights from Conformational Ensembles via Machine Learning

“…ML methods have gained enormous interest in recent years and are now applied in a wide range of research areas within biology, medicine, and health care ( 1 , 2 , 3 ) such as genomics ( 4 ), network biology ( 5 ), drug discovery ( 6 ), and medical imaging ( 7 , 8 ). In molecular simulations, such methods have, for example, eminently been used to enhance sampling by identifying CVs or the intrinsic dimensionality of biomolecular system in a data-driven manner ( 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 ), as an interpolation or exploratory tool for generating new protein conformations ( 23 , 31 , 32 ), as well as providing a framework for learning biomolecular states and kinetics ( 11 , 33 , 34 ). However, many tools borrowed from ML, most notably nonlinear models such as neural networks (NNs), are criticized for their resemblance to a black box, which obstructs human-interpretable insights ( 1 , 2 , 3 , 5 ).…”