Machine Learning Driven Optimization of Fe-Based TMCs for Photodynamic Therapy

Abstract

Noble metal-based photoactive complexes have applicationsin photodynamic therapy (PDT), buttheir toxicity and high cost drive interest in sustainableand cheaper alternatives like iron-basedcompounds. In this paper, quantum chemistry andclassical molecular dynamics were employed tocharacterize the photophysical properties and noncovalentinteractions with DNA of two Fe(III) complexes.We explained the absorption of IR wavelengthby bright ligand-to-metal transitions andshowed that the complexes exhibit persistent, albeitmodest, interaction with DNA. Building onthese traditional simulation methods, we proposea conceptual ML-driven optimization module designedto refine the structure of iron complexes andenhance their photophysical features. While theframework is not yet implemented, we demonstratethat key properties relevant for PDT can be computationallyevaluated, providing a foundation forfuture iterative optimization. The ML module integrates3D molecular structures, simulation results,and quantum chemical insights to suggest modificationsaimed at shifting the absorption spectrummore favorably into the visible range, improvingtheir suitability for phototherapies.