Site of Metabolism Prediction using Graph Neural Networks

Abstract

Quals talk:

Sites of Metabolism (SOMs) are the specific locations within molecules where chemical reactions occur. Because such reactions form the foundation of all biological life, the placement of SOMs has wide-reaching implications in many areas, from drug development to metabolic engineering. A number of SOM prediction methods have been developed, however, they rely on complex sets of features and need to be tailored to specific enzyme types. In this talk, I will present a technique for SOM prediction using Graph Neural Networks (GNNs). This approach takes advantage of graph structure inherent in molecules and is able to learn effective representations of atoms and bonds for SOM prediction across all categories of enzymes. In addition, I will discuss two metabolic engineering applications where SOM prediction with GNNs measurably enhances the quality of results.

Please join meeting in Cummings #601.