These are preliminary reports that have not been peer-reviewed. They should not be regarded as conclusive, guide clinical practice/health-related behavior, or be reported in news media as established information. For more information, please see our FAQs.
Predicting Enzymatic Reactions with a Molecular Transformer
preprintsubmitted on 29.10.2020, 16:10 and posted on 30.10.2020, 12:35 by David Kreutter, Philippe Schwaller, Jean-Louis Reymond
The use of enzymes for organic synthesis allows for simplified, more economical and selective synthetic routes not accessible to conventional reagents. However, predicting whether a particular molecule might undergo a specific enzyme transformation is very difficult. Here we exploited recent advances in computer assisted synthetic planning (CASP) by considering the Molecular Transformer, which is a sequence-to-sequence machine learning model that can be trained to predict the products of organic transformations, including their stereochemistry, from the structure of reactants and reagents. We used multi-task transfer learning to train the Molecular Transformer with one million reactions from the US Patent Office (USPTO) database as a source of general chemistry knowledge combined with 32,000 enzymatic transformations, each one annotated with a text description of the enzyme. We show that the resulting Enzymatic Transformer model predicts the products formed from a given substrate and enzyme with remarkable accuracy, including typical kinetic resolution processes.