Introduction
Polymeric long-acting injectables (LAIs) enhance chronic disease treatment by prolonging therapeutic duration, reducing dosing frequency, and improving patient compliance [1, 2]. Evaluating drug release kinetics—particularly burst release and sustained release—is critical for LAI quality assessment, yet traditional methods face limitations, necessitating predictive modeling [2]. Key challenges include functional data prediction and datasets constraints, which impact model performance [3].
Methods
A systematic review compiled in vitro release studies of polymeric microspheres, extracting key features to construct a database. Missing data (~10%) were imputed using the k-nearest neighbors algorithm. Three machine learning models were trained to predict the complex, non-linear, and high-dimensional release data: classification and regression tree, random forest, and XGBoost. Models were optimized through feature engineering and evaluated via cross-validation, using AIC, R², and MAE. The predictive accuracy of the developed models was further validated in prospective studies.
Results
Over 1,500 in vitro release profiles with >30 numerical and categorical input features were analyzed. Model comparisons showed random forest has the most promising performance, identifying drug/polymer molecular weight, particle size, and drug load as key determinants. Surprisingly, release study conditions (e.g., sampling volume, separation methods) also significantly influenced drug release kinetics. Prospective testing with commercial and in-house formulations confirmed model robustness.
Discussion
This study presents a machine learning framework to predict polymeric LAI release profiles, addressing limitations of traditional methods. The model provides mechanistic insights into release behavior and demonstrates robust predictive accuracy, potentially accelerating formulation development and quality assessment.
Reference
1. Abdelkader, H., et al., Polymeric long-acting drug delivery systems (LADDS) for treatment of chronic diseases: Inserts, patches, wafers, and implants. Advanced Drug Delivery Reviews, 2021. 177: p. 113957.
2. Bao, Q., et al., In vitro release testing method development for long-acting injectable suspensions. Int J Pharm, 2022. 622: p. 121840.
3. Bannigan, P., et al., Machine learning models to accelerate the design of polymeric long-acting injectables. Nature Communications, 2023. 14(1): p. 35.