Determining the conformation of a ligand that promotes binding to a protein active site is one of the most important problems in designing a new drug. In order to be an effective drug, a molecule must have the proper geometry to access the binding site of the target protein. Typically universal force fields are used to determine the molecular conformations, but these methods suffer from limited accuracy. Quantum mechanical (QM) approaches can provide accurate structures and energies, but are too computationally costly to model most systems or to carry out conformational searches in large drug databases. Machine learning (ML) methods offer to bridge the gap between classical and quantum methods by building potential energy surfaces empirically trained to quantum mechanical reference data. The ANAKIN-ME method is used to develop a universal empirical atomistic potential for organic molecules. ANI potentials have the advantage of speeds similar to classical force fields while maintaining chemical accuracy to the QM reference calculation in a transferable and extensible way. ANI potentials are developed to accurately predict not only energies, but also conservative molecular forces since they are the analytic gradient of the potential. Given molecular forces, the calculation of normal mode frequencies and geometry optimization become trivial. We provide test results for several thousand molecules including GDB molecules, tripeptides, and drugbank database molecules. Based on these results, it can be seen that ANI out performs classical and semi-empirical QM methods and has a small error compared to DFT, with a speed comparable to classical methods. ANI accomplishes this without any direct force training and while maintaining a universality in the systems it can be applied to.

Determining molecular normal modes and geometry optimization using the deep learning method ANAKIN-ME

255th National Meeting (2018)

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