Mechanism of Action (MOA) elucidation is drawing increasing interest due to the broad utilization of phenotypic screens as well as a growing interest in toxicity, kinase-selectivity, and poly-pharmacology. Formally, we define MOA as the binding affinity pattern of a small molecule over the proteome or a finite set of proteins of interest. Hence, a powerful MOA algorithm would be able to accurately predict these bioactivity patterns. In this work we investigate AtomNet, the first structure-based convolutional neural network for bioactivity predictions, for its suitability in MOA elucidation. We describe an internal MOA benchmark we are developing and discuss evaluation methodologies for comparing different algorithms for this task, highlighting challenges and pitfalls.

Mechanism-of-action elucidation using deep convolutional neural networks

256th National Meeting (2018)

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