When Nanoparticles Talk to Proteins: Learning the Rules of Nanoscale Recognition

Abstract:
Whether a nanoparticle becomes a therapeutic, a diagnostic, or a toxin is decided at the moment it contacts a biomolecule. Our group has spent years developing molecular-scale descriptions of how nanoparticles form, what chemical identities they acquire, and how those identities govern their behavior in various environments. Yet predicting where and how a given contact occurs remains slow and largely empirical: experiments are costly, atomistic simulations are computationally prohibitive, and existing computational models are built around protein-specific features that do not transfer to engineered nanomaterials.

I will first show what this molecular-scale view makes possible by examining one system in detail. Bacterial biofilms resist both antibiotics and mechanical disruption, in part because their extracellular matrix is reinforced by amyloid fibers assembled from phenol-soluble modulin peptides. Combining simulation with experiment, we found that graphene quantum dots form supramolecular complexes with these peptides, docking near the peptide N-terminus, shifting the peptide's secondary structure away from the alpha-helix, and preventing fibrillation, thereby dispersing mature Staphylococcus aureus biofilms without killing the individual bacteria.

Identifying such interactions one simulation at a time, however, does not scale. I will then describe NeCLAS, a machine learning framework that predicts partner-specific nanoscale interactions across chemical domains. By combining a coarse-grained structural representation with features that encode the physicochemical neighborhood at multiple scales, NeCLAS predicts protein-nanoparticle contacts more accurately than existing methods, remains competitive with protein-specific models for protein-protein interactions, and extends to nanoparticle self-assembly. Together, these efforts point toward rational, predictive design of nanomaterials for biological targets, and our ongoing work pursues both directions: deeper mechanistic understanding of nanoparticle interactions with bacterial communities, and broader predictive models that make such interactions findable by design.

Bio:
Angela Violi is the Dennis Assanis Collegiate Professor and Arthur F. Thurnau Professor at the University of Michigan, where she serves as Associate Chair for Research in Mechanical Engineering. She holds appointments in Mechanical Engineering, Chemical Engineering, Electrical Engineering and Computer Science, Applied Physics, and the Biophysics Program. Her research sits at the intersection of nanoscience, molecular simulation, and biology, using computational methods (from quantum chemistry and molecular dynamics to machine learning) to understand how molecular-scale processes determine the formation, transformation, and function of nanomaterials, and what happens when those materials encounter living systems.

Violi received her M.Sc. and Ph.D. in Chemical Engineering from the University of Naples "Federico II" in Italy. Before joining the University of Michigan faculty in 2006, she held research appointments at the University of Utah and at a Department of Energy ASCI center for science-based simulation of accidental fires and explosions. Her early work made foundational contributions to computational nanoscience, including multiscale models of carbonaceous nanoparticle inception and growth, as well as studies of the health effects of combustion-generated particles. In 2019, she led a U-M BlueSky Initiative that brought together engineering, medicine, microbiology, chemistry, and computational science to pursue nanoparticle design for health applications, the direction that anchors her current work.

Her many honors include the U-M College of Engineering Research Excellence Award, selection as Michigan's nominee for the Michigan Distinguished Professor of the Year competition, the ASME George Westinghouse Silver Medal, the NSF CAREER Award, and election as a Fellow of the Combustion Institute.