Objective

Compound effects can be predicted from structural features using traditional methods, as well as novel approaches derived from the artificial intelligence field. From operationalizing tree-based methods to streamline compound profiling, to novel generative models, there are a number of important contributions of machine learning to drug discovery. In addition to elucidating the activity of known active chemotypes, we explore dark chemical matter and find that it is not inert but rather specific. Finally, we apply machine learning methods to the interpretation of chemogenomic data, as an unbiased approach to uncovering the phenotypic profile of chemical matter.