Being able to accurately predict reaction conditions would be very beneficial for synthetic chemistry and drug development, however it remains a challenge. Some models that have been developed but are slow obviously pose problems when one wants to predict hundreds of millions of reactions. In the context of our Synthetically Accessible Virtual Inventory (SAVI) project, in whose first round we generated more than 283 million novel molecules and their proposed single-step synthetic routes from commercially available starting materials, we will discuss the development of methods to better predict the conditions of these reactions. We will illustrate the use of data from large reaction databases, such as Elsevier Reaxys, to create reaction condition prediction models for SAVI by utilizing machine learning approaches.

Predicting reaction conditions for computer-generated SAVI reactions by machine learning from reaction databases

256th National Meeting (2018)

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