Molecular modeling with machine-learned universal potential functions
This work addresses molecular modeling for drug discovery, presenting an incremental improvement through automated training of neural network-based potential functions.
The paper tackled the problem of molecular modeling for drug discovery by training neural networks as universal approximators for energy potential functions, achieving predictive potential functions on large-scale crystal structures with demonstrated superiority and versatility.
Molecular modeling is an important topic in drug discovery. Decades of research have led to the development of high quality scalable molecular force fields. In this paper, we show that neural networks can be used to train a universal approximator for energy potential functions. By incorporating a fully automated training process we have been able to train smooth, differentiable, and predictive potential functions on large-scale crystal structures. A variety of tests have also been performed to show the superiority and versatility of the machine-learned model.