Machine learning provides predictive analysis into silver nanoparticle protein corona formation from physicochemical...

Area of Science:

• Nanomaterial science
• Biochemistry
• Environmental science

Background:

• Engineered nanomaterials (ENMs) interact with biological and environmental systems.
• Proteins bind to ENMs, forming a protein corona (PC) that alters ENM behavior.
• Predicting PC composition is complex due to protein diversity and varying ENM properties.

Purpose of the Study:

• To develop a predictive model for protein corona populations on ENMs.
• To relate protein, ENM, and solution characteristics to PC formation.
• To support the development of mechanistic models for PC prediction.

Main Methods:

• Utilized random forest classification to model PC formation.
• Related biophysicochemical characteristics of proteins, ENMs, and solution conditions to PC composition.
• Validated the model using experimental data for Ag ENM systems.

Main Results:

• Achieved a model with high performance (AUC=0.83, F1-score=0.81).
• Identified key predictive variables including protein pI, weight, ENM size, surface charge, and solution ionic strength.
• Demonstrated the model's applicability to Ag ENM systems with varying size and coatings.

Conclusions:

• The developed model provides robust predictions of PC fingerprints.
• Protein biophysical properties and ENM characteristics are crucial for accurate PC prediction.
• The model can be adapted for other ENM-PC systems, advancing predictive capabilities.