Prediction of hydrolysis pathways and kinetics of sulfamethoxazole: A machine-learning-based molecular dynamics and experimental study.

Sulfonamide antibiotics (SAs), containing several hydrolytic and ionizable groups, have been included in China's "List of Key Controlled New Pollutants (2023 Edition)". Elucidating hydrolysis pathways and kinetics of SAs is crucial for assessing their persistence in the environment. Herein, sulfamethoxazole (SMX), a widely used and detected SA, was chosen as a case. Based on ab initio molecular dynamics (AIMD) data from reaction systems of SMX and its dissociated forms (i.e., SMX, and SMX), universal machine-learning force field (MLFF) models were trained for SAs, which were validated by SMX and other 11 SAs (with the same atoms as SMX). Furthermore, a series of machine-learning-based molecular dynamics (MLMD) simulations were performed to investigate the possible hydrolysis pathways of SMXs (SMX, SMX and SMX), which were confirmed ca. 60 times faster than the AIMD simulations, with comparable computational precision. The hydrolysis products of SMX were determined using the liquid chromatography-tandem mass spectrometry system, which are consistent with the results from MLMD. Simulations revealed that nearby water molecules can substantially accelerate hydrolysis of SMXs via stabilizing transition states by favorable hydrogen-bonding interaction - a key interaction overlooked by studies relying on widely-used implicit solvent models. This study overcomes the limitations of molecular dynamics simulations and quantum chemical calculations, paving an avenue for predicting emerging pollutant hydrolysis pathways and kinetics in the aquatic environment.