Abstract
The rapid expansion of chemical space presents significant challenges in identifying novel ligands for drug targets. Here, we introduce BIOPTIC B1, an ultra-high-throughput ligand-based virtual screening system capable of rapidly evaluating multi-billion-molecule libraries in minutes. In retrospective analyses across seven diverse drug targets, BIOPTIC B1 demonstrated performance comparable to state-of-the-art machine learning models, including Chemprop and Gradient Boosting. Prospectively, we applied BIOPTIC B1 to leucine-rich repeat kinase 2 (LRRK2), a high-priority drug target for Parkinson's disease. BIOPTIC B1 successfully identified multiple novel ligands binding both wild-type and G2019S-mutant LRRK2 with dissociation constants (K(d)) as low as 110 nM, directly from the Enamine REAL Space of 40 billion molecules. These findings highlight BIOPTIC B1 as a powerful tool for novel hit identification and scaffold hopping within ultra-large chemical space, offering significant advancements in virtual screening for drug discovery.