Multitask Machine Learning of Collective Variables for Enhanced Sampling of Rare Events

• DOI: 10.1021/acs.jctc.1c00143
• Publication Date: 2022-03-11T16:34:58Z
• Authors: Lixin Sun; Jonathan Vandermause; Simon Batzner; Yu Xie; David Clark; Wei Chen; Boris Kozinsky
• Keywords: Adaptive sampling; Chemical space

Computing accurate reaction rates is a central challenge in computational chemistry and biology because of the high cost free energy estimation with unbiased molecular dynamics. In this work, data-driven machine learning algorithm devised to learn collective variables multitask neural network, where common upstream part reduces dimensionality atomic configurations low dimensional latent space separate downstream parts map predictions basin class labels potential energies. The resulting shown be an effective low-dimensional representation, capturing progress guiding umbrella sampling obtain landscapes. This approach successfully applied model systems including 5D Müller Brown model, three-well alanine dipeptide vacuum, Au(110) surface reconstruction unit reaction. It enables automated reduction for controlled reactions complex systems, offers unified data-efficient framework that can trained limited data, outperforms single-task approaches, autoencoders.