Predictive Modeling of Drug Release Kinetics using Supervised Machine Learning Algorithms

Dr. Amar B. Deshmukh ¹, Dr. Mayuri Tushar Deshmukh ², Yugendra D. Chincholkar ³, Dr. A. S. Patil ⁵, Dr. Altaf Osman Mulani ⁶

¹Associate Professor, Department of Electronics & Telecommunication, Anantrao Pawar College of Engineering & Research, Pune, Maharashtra, India.

²Assistant Professor, Department of Electronics & Telecommunication, Marathwada Mitra Mandal College of Engineering, Pune, Maharashtra, India.

³Associate Professor, Department of Electronics & Telecommunication, Sinhgad College of Engineering, Pune, Maharashtra, India.

⁴Associate Professor, Shah and Anchor Kutchhi College of Engineering, Mumbai, Maharashtra, India.

⁵Principal, Bhivarabai Sawant College of Engineering and Research, Pune, Maharashtra, India.

⁶Professor, Department of Electronics & Telecommunication, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India.

---

ABSTRACT

Drug release kinetics is a critical determinant of the therapeutic efficacy and safety profile of pharmaceutical formulations. Traditional empirical and mechanistic modeling approaches, while foundational, are limited in their capacity to capture the complex, multi-dimensional interactions governing drug release from diverse delivery systems. This review paper presents a systematic investigation of supervised machine learning (ML) algorithms for predictive modeling of drug release kinetics, encompassing methodologies from linear regression and support vector machines to ensemble methods and deep neural networks. Drawing upon key peer-reviewed studies published between 2020 and 2026, we critically evaluate the performance metrics, feature engineering strategies, dataset requirements, and validation frameworks employed across polymer-matrix, lipid-based, nanoparticulate, and hydrogel drug delivery platforms. Our analysis demonstrates that gradient boosting algorithms (XGBoost, LightGBM) and hybrid deep learning architectures consistently yield superior predictive accuracy (R² > 0.96) compared to classical kinetic models, particularly for controlled-release and stimuli-responsive systems. We further examine explainability techniques such as SHAPLEY values and LIME, which are increasingly critical for regulatory acceptance of ML-driven formulation workflows. This paper also identifies key challenges including data heterogeneity, limited dataset sizes, and the lack of standardized experimental protocols, and outlines a forward-looking agenda for integrating physics-informed neural networks and automated machine learning (AutoML) into next-generation pharmaceutical development pipelines.

Keywords: Drug release kinetics, supervised machine learning, random forest, XGBoost, neural networks, pharmaceutical formulation, controlled release, SHAP, feature importance, IVIVC.

How to cite this article: Deshmukh AB, Deshmukh MT, Chincholkar YD, Patil AS, Mulani AO. Predictive modeling of drug release kinetics using supervised machine learning algorithms. Int J Drug Deliv Technol. 2026;16(3s): 978-986; DOI: 10.25258/ijddt.16.3s.118

Source of support: None.

Conflict of interest: None