Ml speeds up calculating atomic energy changes in large molecules
Colour me shocked: Exact Molecular Hessians from local MLIPs in O(N) time using sparse differentiation!
Computational Engineering, Finance, and Science
Summary
Calculating how atoms in a molecule interact and change shape usually takes a lot of time, especially for big molecules. The authors figured out a way to use machine learning to drastically speed up these calculations without losing accuracy. Their method works even on large molecules like proteins, which were difficult to handle before. This can help scientists study complex molecules more efficiently.
What this means in practice
- •For computational chemists: Compute precise molecular vibrational properties and energy landscapes for large molecules with significantly reduced computation time using local MLIPs.
- •For drug design teams: Analyze large protein conformations efficiently to guide drug candidate optimization by quickly obtaining accurate molecular Hessians.
Authors
Luca Thiede, Andreas Burger, Alán Aspuru-Guzik
Abstract
The Hessian of the energy with respect to the nuclear positions is indispensable in atomistic modelling. However, constructing this matrix requires $O(N)$ Hessian vector products, traditionally limiting high-accuracy Hessians to small systems. Machine learning interatomic potentials (MLIPs) have accelerated atomistic modelling by providing highly accurate energies and forces at $O(N)$ cost, yet the resulting $O(N^2)$ cost of Hessians remains a practical bottleneck for large systems. Based on the insight that we can derive the sparsity pattern for an MLIP's Hessians in closed form, we show in this paper how to use techniques from sparse automatic differentiation to reduce the cost of a local MLIP's Hessians to a system-size-independent number of Hessian-vector products, yielding overall $O(N)$ total cost without any approximations. We benchmark our approach on a variety of systems ranging from alkane chains to water clusters to $A\beta40$ conformers. Depending on the MLIP configuration, we achieve the linear scaling regime already on relatively small systems, resulting in large runtime reductions between 2$\times$-15$\times$ for these systems. This opens up the possibility of scaling high-accuracy MLIP Hessians to very large systems, such as proteins that were previously inaccessible.