Collaborative Research: Machine Learning for the Protein Corona: An Integrated, Feature-Driven Approach to Predict Nano-Bio Interactions

Nanoparticles are increasingly used in medicine, materials, and agriculture. These nanoparticles come into contact with humans during manufacturing or during use by consumers. In any biological system, proteins adsorb on the surface of nanoparticles forming a coating of proteins on the surface of the nanoparticle, often referred to as a protein “corona.” The specific proteins that adsorb on the nanoparticle surface determine the subsequent interactions of the nanoparticles with cells.

Understanding the nanoparticle properties that influence the protein corona is essential for determining the toxicity associated with human exposure to nanoparticles and developing new nanomedicines and nanosensors. This research aims to predict protein-nanoparticle interactions based on nanoparticle and protein properties using machine learning combined with mechanistic biophysical experiments.

Understanding protein-nanoparticle interactions is vital for industrial and environmental nanoparticle exposures, as well as for therapeutic and diagnostic applications. In addition, this research provides an ideal training platform for students to address fundamental questions of nanoscience using machine learning, providing training relevant to future academic or industry jobs.

This research project aims to predict which proteins will adsorb on the surface of nanoparticles and train students in a highly interdisciplinary environment. The research team will first characterize the protein corona as a function of nanoparticle properties and develop a machine-learning workflow for prediction. The team will vary nanoparticle core composition, ligand, diameter, zeta potential, surface area, and hydrophobicity to sample a wide parameter space.

Proteomics will be used to characterize the adsorption of serum proteins on the nanoparticles. The team will utilize a set of controlled protein features and biophysical assays (isothermal titration calorimetry and nuclear magnetic resonance) to test the predictions from machine learning. Well-defined protein classes will be used to determine whether corona behavior follows expected predictions made by machine learning.

The team will then extend these studies by probing the robustness of machine learning predictions. Challenging mixtures of proteins will be tested, and the observed nanoparticle coronas will be compared to predictions obtained using optimized algorithms. The outcomes of this research will include the proteomics data (shared through ProteomeXchange), machine learning algorithms (shared on GitHub), and a template for recruiting and mentoring first-generation/low-income undergraduate researchers.

TThe ability to predict protein-nanoparticle interactions based on nanoparticle properties will promote the development of nanoparticles for a range of applications and help to determine safe exposure limits.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.