Biomolecular Adsorption on Nanomaterials: Combining Molecular Simulations with Machine Learning

Area of Science:

• Computational chemistry
• Materials science
• Biophysics

Background:

• Understanding biomolecule-nanomaterial interactions is crucial for applications in medicine and environmental science.
• Predicting adsorption behavior is complex due to the variety of biomolecules and nanomaterials.

Purpose of the Study:

• To analyze adsorption free energies of small biomolecules on various nanomaterials.
• To develop predictive machine learning (ML) models for these interactions.
• To establish a method for grouping nanomaterials based on biomolecule interactions.

Main Methods:

• Computed adsorption free energies using molecular dynamics-metadynamics.
• Applied unsupervised learning (principal component analysis, clustering) and supervised learning (regression, neural networks).
• Developed ML models to predict adsorption energies.

Main Results:

• Identified a core set of biomolecules whose adsorption energies can predict others.
• Developed ML models capable of predicting adsorption free energies.
• Presented a methodology for classifying nanomaterials by their interaction profiles with biomolecules.

Conclusions:

• Machine learning effectively models biomolecule-nanomaterial adsorption.
• A reduced set of biomolecules can serve as surrogates for predicting adsorption.
• The study provides a framework for understanding and predicting complex adsorption phenomena.