Learning to Steer Eco-coronas: An AI-driven Framework for Protein-guided Control of Environmental Nano–bio-interfaces

The formation of eco-coronas is a key issue that must be understood and predicted to ensure the safety and environmental performance of nanomaterials. Eco-coronas are biomolecular coatings that form on the surface of nanoparticles in environmental media, where aggregation behavior, bioavailability, and ecological fate are influenced by proteins, lipids, and microbial metabolites. Predictive and mechanistic control remains limited, despite previous studies qualitatively characterizing eco-corona composition. In this study, a hybrid framework, EcoCorona-MAP (Model–Actuate–Protect), is presented as an artificial intelligence-based system that forecasts and guides eco-corona formation. The system is based on a graph neural network (GNN) trained on environmental proteomic data from Danio rerio and Daphnia magna microbiomes to predict protein–nanoparticle adsorption energies (regression task). Predicted adsorption energies are subsequently used to estimate adsorption probabilities, which are used to design decoy peptides that competitively bind to nanoparticle surface sites to modulate eco-corona composition. Experimental validation using TiO₂ and carbon nanotube (CNT) nanoparticles revealed that approximately 40% of microbial protein adsorption was reduced and colloidal persistence was enhanced under simulated aquatic conditions. Simulations also indicated that peptide–nanoparticle binding remained stable, as confirmed by molecular dynamics simulations (RMSD < 2.5 Å). Overall, EcoCorona-MAP establishes a predictive-to-interventional framework that integrates artificial intelligence with molecular design to actively control nano–bio interfaces, providing a foundation for environmentally responsive and safer-by-design nanomaterials.