Organic chemistry serves as the basis for life, making it the backbone in the synthesis of biologically active compounds, pharmaceuticals and therapeutically important substances. A proper computational description of the physics involved in organic reactions will assist in solving many chemical problems. However, traditional force fields based on classical principles are unable to model bond dissociations, while quantum mechanical approaches are computationally too intense. Machine learned (ML) potentials have recently shown to be accurate in predicting reaction profiles. However, these ML potentials are highly specialized tools, which are trained to solve a particular task, lacking an “out-of-the-box” level of transferability. To overcome these challenges, we develop the necessary tools and data set based on the concept of active learning to train ANAKIN-ME (ANI), a deep learned atomistic potential to predict organic reaction profiles with an accuracy of DFT at the computational cost of force field. The new ANI potential demonstrates reusing existing knowledge in determining energetics for the types of reaction classes it was trained on, extending low error to systems well outside of the training data set. With the enhancement of chemical space covered, while maintaining universality in the systems it can be applied to, ANI can be a powerful tool to assist organic chemists and a game changing paradigm for drug design.<br/>

Transference of knowledge in deep learning for chemically accurate organic reaction profiles

2019 Spring National Meeting

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