Meta-Analysis and Machine Learning Prediction of Protein Corona Composition across Nanoparticle Systems in Biological...

Area of Science:

• Biomaterials Science
• Nanotechnology
• Proteomics

Background:

• Protein corona (PC) formation on nanoparticles (NPs) critically impacts their biological fate and therapeutic efficacy.
• Current understanding is limited by a lack of standardized, comprehensive datasets on PC composition.
• Systematic analysis requires integrating data from diverse NP formulations and experimental conditions.

Purpose of the Study:

• To introduce the Protein Corona Database (PC-DB), a centralized resource for NP-protein interactions.
• To analyze the heterogeneity of PC composition across various NP types and physicochemical properties.
• To identify key NP characteristics that predict specific protein adsorption patterns.

Main Methods:

• Compilation of data from 83 studies (2000-2024) into the PC-DB, integrating 817 NP formulations and 2497 adsorbed proteins.
• Meta-analysis of NP materials, surface modifications, sizes, and ζ-potentials.
• Application of interpretable machine learning models (LightGBM, XGBoost) to predict protein adsorption.

Main Results:

• PC-DB reveals significant heterogeneity in NP materials (metal, silica, lipid-based), sizes (1-1400 nm), and ζ-potentials (-70 to +70 mV).
• Smaller NPs (<100 nm) with neutral/negative ζ-potentials preferentially bind lipoproteins (APOE, APOB-100), enhancing delivery.
• Negatively charged metal NPs bind complement C3, suggesting increased immune clearance.
• NP size, ζ-potential, and incubation time are key predictors of protein adsorption (ROC-AUC > 0.85).

Conclusions:

• Physicochemical properties of NPs are major determinants of protein corona composition.
• The PC-DB provides a valuable resource for understanding NP-protein interactions.
• Predictive modeling can guide the rational design of NPs for optimized therapeutic outcomes.