Machine Learning for Predicting Drug Release Behavior of PLGA Microspheres

Area of Science:

• Biomaterials Science
• Pharmaceutical Technology
• Computational Chemistry

Background:

• Poly(lactic-co-glycolic acid) (PLGA) microspheres are crucial for sustained drug delivery, enhancing patient compliance.
• Current development relies on inefficient trial-and-error methods due to complex formulation factors influencing drug release.

Purpose of the Study:

• To develop and validate machine learning models for predicting in vitro drug release profiles from PLGA microspheres.
• To identify key formulation parameters influencing drug release kinetics.

Main Methods:

• A comprehensive dataset of 113 PLGA formulations was compiled from scientific literature.
• Multiple machine learning algorithms were trained and evaluated for predictive performance.
• Feature importance analysis was performed to understand critical factors affecting release.

Main Results:

• The optimal machine learning model achieved high accuracy, with R² = 0.9415, RMSE = 6.99%, and MAE = 4.35%.
• The model demonstrated robust predictive capability for in vitro drug release from PLGA microspheres.
• Key factors influencing drug release were identified through feature importance analysis.

Conclusions:

• Machine learning provides a powerful tool for the rational design and optimization of PLGA-based drug delivery systems.
• Accurate prediction of drug release profiles can accelerate formulation development and improve therapeutic outcomes.
• This data-driven approach facilitates the efficient development of long-acting injectable formulations.