Predicting tissue distribution and tumor delivery of nanoparticles in mice using machine learning models

Area of Science:

• Nanomedicine
• Computational Biology
• Materials Science

Background:

• Targeted delivery of nanoparticles (NPs) in cancer nanomedicine faces challenges with low delivery efficiency (DE) to tumor sites.
• Understanding how NP physicochemical properties affect tissue distribution and tumor DE is crucial for improving nanomedicine design.

Purpose of the Study:

• To develop and validate machine learning models for predicting NP tissue distribution and tumor DE.
• To identify key NP physicochemical properties influencing delivery efficiency.
• To create a computational tool for high-throughput screening of nanomedicine formulations.

Main Methods:

• Trained and validated multiple AI models, including deep neural networks (DNNs), using the Nano-Tumor Database.
• Evaluated model performance using determination coefficients (R²) and root mean squared error (RMSE).
• Conducted feature importance analysis to identify critical NP properties for delivery prediction.

Main Results:

• The DNN model demonstrated superior prediction accuracy for tumor DE and major tissue distribution compared to other models.
• High R² values were achieved for predicting DE in tumors (0.41) and various organs (e.g., spleen 0.79, lung 0.87, kidney 0.83).
• NP core material was identified as a significant factor influencing delivery predictions.

Conclusions:

• The developed DNN model serves as an effective high-throughput pre-screening tool for designing efficient cancer nanomedicines with enhanced tumor DE.
• This computational approach can help reduce, refine, and partially replace animal experimentation in nanomedicine research.
• The model's conversion to a web dashboard facilitates broader application in nanomedicine development.