ML-assisted Computational Workflow to the Robust Modeling of Supramolecular Metallacages in Solution

We present here an integrated computational workflow to study the behavior of selected self-assembled [Pd2L4]4+ metallacages in explicit solvents (water and DMSO) and benchmark them for future in silico investigation of their host-guest chemistry, pivotal to their application as drug delivery systems. The workflow includes classical force field (FF) construction of two metallacage analogues and testing for standard and Machine Learning Interatomic Potential (MLIP)-based molecular dynamics (MD) simulations. Overall, the study provides a direct comparison of classical FF with MILPs with respect to performance vs computational cost. The proposed workflow offers a versatile framework for computationally screening and designing supramolecular systems in solution, effectively bridging the gap between quantum-level accuracy and the temporal and spatial scales needed for simulations of different functional applications.