Fig. 3

Pipeline for archived and virtual libraries searching. Modules in the pipeline and the parameter cutoffs are illustrated. The process is: diverse hit scaffolds from the output after two layers filtering and SAR enrichment; Shape and electrostatics ROCS searching in libraries; GBVI/WSA ΔG calculation; more optional filters can be added such as compound amount, compound QC results, solubility experimental data or machine learning (ML) prediction, and bioactivity machine learning model prediction, to generate a focused set of compounds for biological testing