AI-Driven Discovery of Highly Specific and Efficacious hCES2A Inhibitors for Ameliorating Irinotecan-Triggered Gut Toxicity

Abstract

The anticancer agent irinotecan often induces severe delayed-onset diarrhea, inhibiting human carboxylesterase 2A (hCES2A) can significantly alleviate irinotecan-triggered gut toxicity (ITGT). This work presents an efficient workflow for de novo design and developing novel efficacious hCES2A inhibitors. A well-training machine learning model identified scaffold-14 as a lead compound, while compound 14n was developed as a novel time-dependent hCES2A inhibitor (IC₅₀ = 0.04 nM) following three rounds of structural optimization. The covalent binding modes and inactivation mechanisms of 14n were elucidated by nanoLC-MS/MS-based chemoproteomics and covalent docking simulations. Notably, 14n showed excellent selectivity, good cell-membrane permeability, favorable drug-like properties, and potent inhibition on intracellular hCES2A. In vivo tests demonstrated that 14n was orally active, showing favorable safety profiles and impressive ameliorative effects on ITGT in tumor-bearing mice. Collectively, this work showcases a high-efficient AI-driven strategy for developing novel efficacious hCES2A inhibitors, while 14n emerges as a promising candidate for alleviating ITGT.