A Generative Adversarial Network Approach to Predict Nanoparticle Size in Microfluidics

Area of Science:

• Materials Science and Engineering
• Biomedical Engineering
• Chemical Engineering

Background:

• Nanoparticles (NPs) are crucial for advanced drug delivery systems (DDS).
• Precise control over NP size in microfluidic synthesis is essential for optimizing DDS performance.
• Poly(lactic-co-glycolic acid) (PLGA) NPs are widely investigated for drug delivery applications.

Purpose of the Study:

• To develop accurate predictive models for poly(lactic-co-glycolic acid) (PLGA) nanoparticle size synthesized via microfluidics.
• To identify key parameters influencing NP size in microfluidic systems.
• To enhance the reliability and accuracy of NP size prediction using machine learning.

Main Methods:

• A comprehensive database of over 1100 data points was curated from extensive literature review.
• Tabular Generative Adversarial Network (TGAN) was utilized for data augmentation to improve dataset reliability.
• Multiple machine learning algorithms, including Decision Tree (DT), Random Forest (RF), Deep Neural Networks (DNN), Linear Regression (LR), Support Vector Regression (SVR), and Gradient Boosting (GB), were employed for NP size prediction.

Main Results:

• The Decision Tree (DT) algorithm demonstrated the highest accuracy in predicting NP size, with an average prediction error of 8%.
• Simulations highlighted the significant impact of synthesis method, poly(vinyl alcohol) (PVA) concentration, and the lactide-to-glycolide (LA/GA) ratio of PLGA copolymers on NP size.
• Data enhancement using TGAN improved the overall prediction accuracy and reliability of the machine learning models.

Conclusions:

• Machine learning, particularly Decision Tree, offers a robust approach for predicting PLGA NP size in microfluidic synthesis.
• Key formulation parameters such as PVA concentration and LA/GA ratio are critical determinants of NP size, enabling better control over DDS characteristics.
• This study provides valuable insights for optimizing microfluidic synthesis of PLGA NPs for enhanced drug delivery applications.