Guiding the rational design of biocompatible metal-organic frameworks for drug delivery

Abstract

Metal-organic frameworks (MOFs) are promising for drug delivery due to their high drug-loading capacity, tunable porosity, and structural diversity. However, their clinical translation is hindered by concerns over biocompatibility. Unfortunately, experiments are resource and time intensive, while modeling approaches fail to capture the behavior of MOFs in intricate biological systems. Herein, we report a machine learning (ML)-guided computational pipeline for probing MOF biocompatibility. Using a database of over 35,000 organic molecules, our interpretable ML models predict the toxicity of MOF linkers with over 80% accuracy across different administration routes. Furthermore, we cataloged the toxicity of MOF metallic centers and screened 86,000 MOFs from the Cambridge Structural Database, identifying candidates with minimal toxicity profiles. Beyond screening, our models provide insights into chemical features of biocompatible MOFs—enabling de novo rational design. This framework expedites the discovery of safer MOFs for drug delivery and deepens the understanding of their underlying chemistry.