Hit identification in ultra large virtual screening: an integrative review and future challenges

Ultra-large virtual screening (ULVS) enables systematic evaluation of hundreds of millions to billions of compounds for hit discovery. This review surveys recent ULVS strategies spanning structure-based, ligand-based, pharmacophore-based, fragment-based and hybrid workflows, often augmented by machine learning (ML) and deep learning. Platforms such as VirtualFlow, RosettaVS, Deep Docking and V-SYNTHES have delivered chemically novel, experimentally validated hits across diverse target classes, including G-protein-coupled receptors and protein-protein interfaces. Despite these successes, challenges remain in scoring accuracy, resource efficiency and generalizability. Future ULVS will increasingly emphasize selective, adaptive exploration through ML-guided prioritization, fragment-based enumeration and synthesis-aware library design.