An extension of jamdock-suite for virtual screening from raw chemical datasets

Virtual screening of chemical libraries has long been a cornerstone for identifying candidate bioactive compounds. Recently, many tools have been developed to make virtual screening more accessible. Among them, the jamdock-suite provides an automated and user-friendly workflow that encompasses the entire process from library generation to docking evaluation. However, its library generation is restricted to the ZINC database and cannot process user-supplied chemical datasets. Here, we present semdock, an extension of the original workflow that enables the preparation of docking-ready ligands from raw SDF, SMILES, and CSV datasets through an automated ligand preparation procedure. The workflow was validated using 3,230 FDA-approved compounds and multiple protein targets. We also implemented ligand-wise CPU-parallel docking, reducing runtime by approximately 30% compared with conventional multithreaded execution. Evaluation across multiple protein targets confirmed that the extended workflow preserves the utility of the original framework while expanding its use to diverse public and user-defined datasets.