Expanding Bioactive Fragment Space with the Generated Database GDB-13s

Identifying innovative drug-like small molecules is critically important in medicinal chemistry to address new targets and overcome limitations of classical molecular series. By deconstructing molecules into ring fragments (RFs, ring and ring-adjacent atoms) and acyclic fragments (AFs, only acyclic atoms), we find that molecules in public databases of drug-like compounds (ZINC, PubChem) and natural products (COCONUT) mostly consist of RFs and AFs up to 13 atoms, and that many RFs and AFs are enriched in bioactive compared to inactive molecules in ChEMBL. We then search the 28,246,012 RFs and 2,640,023 AFs in the generated database GDB-13s (99,394,177 molecules up to 13 atoms following simple functional group and ring strain criteria) for subsets resembling ChEMBL bioactive RFs and AFs. Many of these RFs and AFs are structurally simple, have favorable synthetic accessibility scores, and represent opportunities for synthetic chemistry to contribute to drug innovation in the context of fragment-based drug discovery.