Deciphering Zinc Binding Groups through Systematic Modeling and Cheminformatic Analysis.

Zinc metalloenzymes (Zn-MEs) are a target family with significant therapeutic relevance. Zinc binding groups (ZBGs) are key for Zn-ME inhibitors. However, the limited structural diversity of available ZBGs hampers drug discovery efforts targeting Zn-MEs. To facilitate the identification of novel ZBGs, we attempted to decipher ZBG features via systematic modeling and cheminformatic analysis based on accumulated data related to Zn-MEs as well as their inhibitors. Bioactive ligands targeting Zn-MEs were first collected as the source data set. Twelve binary classifiers with different characterization strategies and algorithms were then applied to distinguish Zn-ME inhibitors from nonmetalloenzyme inhibitors. We subsequently employed multiple interpretable machine learning and substructure detection methods to define ZBGs and accordingly proposed a specific ZBG library containing 41 SMARTS patterns. We applied this library to three different scenarios and exemplified its capability to enrich potential Zn-ME inhibitors, optimize drug-likeness profiles, and modulate subtype selectivity. Given the challenges in identifying novel ZBGs, strategies and methods presented herein are expected to expand the resorts for drug discovery targeting Zn-MEs.