An Explainable Deep Learning Framework with SHAP Analysis for Predicting Drug Release Kinetics in Polymer-Based Controlled Delivery Systems
International Journal of Drug Delivery Technology · 06.05.2026 · 16 (36s)
DOI: 10.25258/ijddt.16.36s.74 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 06.05.2026. Verfasst von Dr. Arpankumar Raval und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Controlled drug delivery systems, particularly PLGA-based polymeric nanoparticles and HPMC matrix tablets, have become widely adopted for delivering sustained and targeted drug actions. However, predicting drug release behavior remains a challenging task due to the nonlinear dynamics involved in the interactions between formulation components and environmental factors. Conventional mathematical models such as the Higuchi equation and Korsmeyer-Peppas model tend to be incapable of handling these complexities, while deep learning models, although highly accurate, lack interpretability. In this paper, an explainable deep learning model is proposed to predict drug release behavior from controlled drug delivery systems. A feed-forward deep neural network is designed to establish a relationship between five key physicochemical formulation parameters — polymer concentration, drug loading, particle size, pH of dissolution medium, and time — and their associated drug release profiles. A dataset comprising 500 experimental formulations was used to train and evaluate the proposed model. To enhance interpretability, SHapley Additive exPlanations (SHAP) are incorporated into the framework. The model achieved an R² of 0.94, RMSE of 3.12%, and MAE of 2.45%, significantly outperforming baseline models including Linear Regression (R²=0.71) and Random Forest (R²=0.86). SHAP analysis revealed that polymer concentration and particle size are the most influential parameters governing drug release behavior, findings that are consistent with established pharmaceutical principles. Not only is the proposed model effective in terms of predictive accuracy, but it also serves as a practical tool for rational formulation design. This approach can significantly reduce experimental burden and accelerate drug delivery system development in the pharmaceutical industry.
Zitieren
APA
Raval, D. A., Rathod, D. G., Trivedi, S. B., Rana, K., & Patel, D. S. (2026). An Explainable Deep Learning Framework with SHAP Analysis for Predicting Drug Release Kinetics in Polymer-Based Controlled Delivery Systems. International Journal of Drug Delivery Technology, 16(36s). https://doi.org/10.25258/ijddt.16.36s.74
Vancouver
Raval DA, Rathod DG, Trivedi SB, Rana K, Patel DS. An Explainable Deep Learning Framework with SHAP Analysis for Predicting Drug Release Kinetics in Polymer-Based Controlled Delivery Systems. International Journal of Drug Delivery Technology. 2026;16(36s). doi:10.25258/ijddt.16.36s.74
BibTeX
@article{Raval2026,
author = {Raval, Dr. Arpankumar and Rathod, Dr. Ghanshyam and Trivedi, Saumil B. and Rana, Khushbu and Patel, Dr. Sanskruti},
title = {An Explainable Deep Learning Framework with SHAP Analysis for Predicting Drug Release Kinetics in Polymer-Based Controlled Delivery Systems},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {36s},
doi = {10.25258/ijddt.16.36s.74}
}
EUVIMED