Deep learning systems now impact our daily lives through voice recognition, image recognition and more recently self-driving cars. Deep-learning has been well studied for categorization and recognition of categories such as recognizing that an image is of liver cells, and that the liver cells are abnormal. Here we present initial analysis of chemical reactions studied by deep learning on the 3D electron densities of reactants and products to predict how molecules might be transformed in chemical reactions. The computational approach is similar to those now used to colorize black and white photos, extended to 3D fields. One advantage of this method over other approaches is that it can easily be applied to inorganic systems which include catalysts. The system is trained on a set of reactions from Reaxys to study the feasibility of the approach. Results and issues will be discussed.

Exploring the use of conditional generative adversarial networks (cGAN) to analyze chemical reactions via electron density fields

256th National Meeting (2018)

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