Machine Learning tools promise to revolutionize drug discovery, greatly expand the set of druggable targets, and have shown unbelievably good performance on benchmarks for Medicinal Chemistry tasks.  Unfortunately, recent analyses imply that such apparently-encouraging algorithmic performance are likely to be an artifact of weaknesses in the design of these benchmarks.  Therefore, we are running the largest assessment of machine learning for hit discovery in history, comprising over 350 projects on 238 targets with 112 universities in 19 countries.  We present the first results from this broad experiment, and report on lessons learned and best practices for medicinal chemists considering artificial intelligence tools and for academic researchers embarking on translational research.

Lessons from the world's largest prospective application of machine learning to hit discovery

2019 Fall National Meeting

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