Identifying factors controlling cellular uptake of gold nanoparticles by machine learning

Area of Science:

• Nanomedicine
• Biotechnology
• Computational Biology

Background:

• Gold nanoparticles (GNPs) show therapeutic promise but require understanding of their biological interactions.
• Cellular uptake of GNPs is variable and not fully understood, hindering safe development.
• Data-driven models can elucidate factors contributing to uptake variability.

Purpose of the Study:

• To develop and compare machine learning models for predicting GNP cellular uptake.
• To identify key nanoparticle properties and experimental conditions driving uptake variability.

Main Methods:

• Trained machine learning models on 2077 data points from 59 studies.
• Evaluated five ensemble learning algorithms: Xgboost, random forest, bootstrap aggregation, gradient boosting, and light gradient boosting machine.
• Identified key predictors of GNP cellular uptake.

Main Results:

• Ensemble machine learning models achieved high prediction accuracy, explaining 80-90% of the variance in GNP uptake.
• Particle size, zeta potential, GNP concentration, and exposure duration were identified as primary drivers of cellular uptake.
• The developed models demonstrate robust predictive capabilities for GNP cellular internalisation.

Conclusions:

• Machine learning effectively models GNP cellular uptake, offering insights into variability.
• This approach can guide the design of GNPs with optimized cellular internalisation for therapeutic applications.
• The study promotes better utilization of existing data and minimizes bias in future research.