Machine-Learning Frameworks for Metal-Ligand Modelling: Applications in Catalysis and Drug Design

Understanding molecular structure, reactivity, and dynamics is essential for advancing catalyst design, drug discovery, and sustainable synthesis. Molecular function arises from the interactions between molecules and their surroundings, which involve a wide range of intermolecular interactions. Among these, metal-ligand interactions stand out due to their tunability. These interactions are key in homogeneous catalysts, supramolecular assemblies, and enzymes. However, accurately modelling them in solution remains challenging, particularly for flexible systems or where solvent effects are relevant. machine learning interatomic potentials (MLIPs) offer a promising avenue to surpass current limitations, but their broad applicability is hindered by challenges in representation, training costs, and transferability.This project introduces transformative approaches for modelling metal complexes in solution, integrating method development, applications, and experimental validation. Specifically, the project will:1.Develop MLIP training strategies to model metal complexes across diverse environments.2.Establish quantitative modelling framework to uncover mechanisms of processes such as self-assembly and speciation; aiding the design of novel structures.3.Explore the origin of catalysis in supramolecular cages using MLIPs and hybrid approaches, guiding the design of novel catalysts with generative models.4.Extend these frameworks to metal-ligand interactions in biological systems to guide the design of novel inhibitors for antimicrobial resistance (AMR).This integrative approach will deliver unparalleled precision in modelling metal complexes and enable breakthroughs in catalysis and drug discovery. Experimental validation and interdisciplinary collaboration will ensure impactful outcomes.