2025 AIChE Annual Meeting

(121i) Remixing Enhanced Sampling Methods in Free Energy Calculations: Alchemical Metadynamics and Exchanges of Expanded Ensembles

Authors

Michael Shirts - Presenter, University of Colorado Boulder
Wei-Tse Hsu, University of Colorado Boulder
Anika Friedman, University of Colorado Boulder
Lindsey Whitmore, University of Colorado Boulder
A large number of enhanced sampling methods have been developed for molecular simulations, but the differences in these many simulation methods often boil down to different arrangements of a limited number of different modular "components". Several interesting approaches can still emerge when one considers the different ways these methodological components can be combined. In particular, we present two different approaches for improving the conformational sampling of alchemical free energy calculation methods with slow degrees of freedom, each combining standard accelerated sampling methods in useful ways. Both approaches expand the range of applicability of the methodological "ingredients" that we can construct methods out of, and each address different challenges that can arise in complex free energy calculations with long-lasting metastable states.

We first present alchemical metadynamics, which allows explicitly biasing configurational collective variables simultaneously with alchemical free energy calculations, entirely within the metadynamics framework. Using test systems with varying complexity, we demonstrate that alchemical metadynamics can enable sampling of configurational metastable states not necessarily accessible in standard free energy methods.

We also introduce the method of replica exchange of expanded ensemble (REXEE), where one which periodically exchanges coordinates of different replicas of expanded ensemble simulations that collectively span the range of interest in the auxiliary variable(s). We show that this method yields free energies consistent with the estimates obtained in expanded ensemble and Hamiltonian replica exchange, but offers significant additional flexibility in how one allocates alchemical states between processors and lessens the need for tight coupling between these processors. It also removes some of the issues that occur with expanded ensemble where state ranges can remain unsampled.

Finally, we present generalizations of REXEE to allow exchange between multiple chemistries, Multiple Topology Replica Exchange (MT-REXEE), and discuss its applications to problems such as enhanced conformational sampling of long alkyl chains via chain growth, the simultaneous sampling of multiple R-group transformations in drug lead optimization, and as a way to easily peform continuous fractional Monte Carlo within molecular dynamics code without modification of the MD code itself.