Discovery of a novel natural USP25 inhibitor via an integrated deep learning-guided screening strategy with molecular dynamics validation.
TL;DR
An integrated workflow coupling AI-based activity prediction and molecular docking to mine natural products as novel USP25 inhibitors is established, serves as an economical and efficient substitute for conventional high-throughput screening and creates a robust technical paradigm for early-stage drug discovery.
Abstract
Ubiquitin-specific protease 25 (USP25) is a critical deubiquitinase implicated in inflammation, neurodegenerative disorders and oncogenic signaling pathways. However, the scarcity of structurally diverse small-molecule inhibitors restricts mechanistic investigation and therapeutic development targeting USP25. In this study, we established an integrated workflow coupling AI-based activity prediction and molecular docking to mine natural products as novel USP25 inhibitors. Through in vitro enzymatic functional validation, theaflavin 3,3'-digallate (TF) was identified as a potent USP25 inhibitor with an IC50 of 33 nM and 8.5-fold selectivity over its paralog USP28. Static molecular docking uncovered abundant hydrogen bonds and π-π stacking interactions between TF and USP25, while 100 ns molecular dynamics simulations and MM-PBSA thermodynamic analysis confirmed stable ligand anchoring. Apart from robust USP25 inhibitory capacity, TF also displayed evident antioxidant activity. Beyond advancing the research of USP25 inhibitors, the computational workflow reported here creates a robust technical paradigm for early-stage drug discovery. For targets short of co-crystal structures and validated lead compounds, this method serves as an economical and efficient substitute for conventional high-throughput screening.