One of the fundamental limitations of accurately modeling biomolecules like DNA is the inability to perform quantum chemistry calculations on large molecular structures. This limitation is particularly relevant when modeling DNA in solution due to the inclusion of large numbers of solvent molecules. We present a machine learning model based on an equivariant Euclidean neural network framework to obtain accurate ab initio electron densities for arbitrary DNA systems in solution that are much too large for conventional quantum methods. The model can be used to study DNA's interactions with solvent at an ab initio level of theory. A promising application for the machine learning model involves calculating forces from machine learned electron densities through the Hellmann-Feynman theorem. The efficient calculation of forces can be useful for algorithms employed in ab initio molecular dynamics, leading to potential applications in biomolecular binding and structural prediction for DNA macromolecules.

Constructing an ab initio DNA and solvent interaction model with equivariant neural networks

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.