ACTIVE LEARNING ON SYNTHONS FOR MOLECULAR DESIGN

Exhaustive screening is highly informative but often intractable against expensive objective functions involved in modern drug discovery. This problem is exacerbated in combinatorial contexts such as multi-vector screening, where molecular spaces can quickly become too large. Here, we introduce Scalable Active Learning via Synthon Acquisition (SALSA): a simple algorithm for efficient multi-vector expansion which extends pool-based active learning to non- enumerable spaces by factoring modeling and acquisition over synthon or fragment choices. Through experiments on ligand- and structure-based objectives we highlight SALSA’s sample efficiency, and its ability to scale to spaces of trillions of compounds. Further, we demonstrate application toward multi-parameter objective design tasks on three protein targets – finding SALSA-generated molecules have comparable chemical property profiles to known bioactives, and exhibit greater diversity and higher scores over an industry-leading generative approach.