The Challenge
Understanding What Drives Metal Oxide NP Toxicity
Metal oxide nanoparticles are among the most widely produced engineered NPs, used in consumer products from sunscreens to electronics. Yet their ability to cross biological barriers and cause toxic effects makes hazard assessment essential for safe use.
Traditional experimental approaches cannot keep pace with the diversity of engineered NPs. Computational modelling offers a high-throughput alternative, but standard physicochemical descriptors alone are insufficient — atomistic-level structural features are needed to capture the mechanisms driving toxicity.
Our Approach
Combining experimental and atomistic data to build interpretable nanotoxicity models
Retrieve and curate toxicological data
Retrieved cytotoxicity data for 24 metal oxide NPs from the S²NANO database, covering ATP and LDH assays on BEAS-2B and RAW 264.7 cell lines, with 15 physicochemical and structural descriptors.
Enrich with 62 atomistic descriptors
Calculated full-particle atomistic computational descriptors based on crystal structure information alone, using molecular mechanics to derive coordination numbers, energetics, force vectors, and surface properties for each NP.
Build QNAR model on Isalos Analytics
Used the Isalos Analytics Platform to develop a quantitative nanostructure-activity relationship model from the combined 77-descriptor feature space, systematically evaluating statistical significance of each descriptor.
Identify key toxicity drivers
Identified 7 statistically significant descriptors driving cytotoxicity: NP core size, hydrodynamic size, assay type, exposure dose, conduction band energy, metal atom coordination number, and average force vector surface normal component.
Deploy and share through H2020 projects
Published the model as a public web service and integrated it into the NanoSolveIT Integrated Approach to Testing and Assessment (IATA) framework, with data available through the NanoPharos database.