Integrated Computational Design and Multistage Virtual Screening for the Prioritization of Putative USP5 Catalytic-Cleft Inhibitors

Ubiquitin-specific protease 5 (USP5) processes unanchored polyubiquitin and has been linked to cancer-associated signaling. We developed a computer-aided workflow to generate and rank small molecules against its catalytic/substrate cleft. The starting dataset contained 26 activity records representing 19 unique compounds, ten with measured half-maximal inhibitory concentrations. Descriptor-based quantitative structure-activity relationship models underwent compound leave-one-out and scaffold-held-out validation. Ten chemical-enumeration methods produced 3,646 method–product records and 3,264 unique structures. The main screen combined these structures with ten additional source-positive molecules and retained 18 compounds under strict selection criteria. A ten-compound panel, comprising nine generated structures and one known benchmark, underwent corrected AutoDock Vina docking against two USP5 receptor copies. The primary and repeated-seed campaigns produced 800 recorded modes. Standard-precision (SP) and extra-precision (XP) Glide docking independently evaluated the prepared ligand library, followed by Prime molecular mechanics/generalized Born surface area (MM-GBSA) calculations on the XP complexes. P06 emerged as the top-ranked candidate in the three-stage consensus analysis, ranking first in SP and second in both XP and MM-GBSA evaluations. P03 achieved the most favorable mean XP docking score, whereas P09 exhibited the strongest mean MM-GBSA binding energy. P10 ranked third overall but demonstrated greater sensitivity to receptor conformation. Collectively, this integrated computational workflow establishes a prioritized hierarchy of candidate compounds targeting the USP5 catalytic/substrate-binding cleft and highlights the influence of chemical selection and receptor preparation on structure-based rankings.