Hybrid Response Surface-Machine Learning Optimization of Doxorubicin-Loaded Chitosan Nanoparticles.
Sukaina Nimrawi1, Nusaiba K Al-Nemrawi2, Young Min Kwon1
1Department of Pharmaceutical Sciences, Barry and Judy Silverman College of Pharmacy, Nova Southeastern University, Fort Lauderdale, Florida 33328, United States.
ACS Applied Bio Materials
|July 15, 2026
Summary
This study introduces a hybrid Response Surface Methodology (RSM) and Machine Learning (ML) approach for optimizing chitosan nanoparticle (CS NP) formulations. This data-efficient strategy accurately predicts critical quality attributes for improved drug delivery systems.
Area of Science:
- * Pharmaceutical Sciences
- * Nanotechnology
- * Materials Science
Background:
- * Chitosan nanoparticles (CS NPs) are promising for drug delivery due to biocompatibility and biodegradability.
- * Traditional formulation optimization methods (OFAT, trial-and-error) are inefficient and fail to capture complex variable interactions.
- * A need exists for advanced, data-driven strategies to optimize CS NP formulations for critical quality attributes (CQAs).
Purpose of the Study:
- * To develop and validate a hybrid Design of Experiments (DoE) strategy integrating Response Surface Methodology (RSM) and Machine Learning (ML) for CS NP formulation.
- * To systematically evaluate formulation variables impacting CS NP CQAs: particle size (PS), polydispersity index (PDI), zeta potential (ZP), and encapsulation efficiency (%EE).
- * To identify the most effective ML algorithm for predicting CS NP formulation performance.
Main Methods:
- * Ionic gelation was used to prepare doxorubicin hydrochloride (DOX HCl)-loaded CS NPs.
- * A DoE strategy combined RSM for modeling and ML algorithms (linear regression, k-NN, decision trees, forests, boosted trees, GPR, ANN, SVM) for prediction.
- * Three formulation variables were optimized to influence PS, PDI, ZP, and %EE.
Main Results:
- * RSM models showed strong statistical adequacy in predicting CS NP CQAs.
- * Machine learning analysis identified Support Vector Machines (SVM) as the superior predictive model across all evaluated CQAs.
- * Experimental validation confirmed close agreement between predicted and observed responses from both RSM and ML models.
Conclusions:
- * The hybrid RSM-ML approach provides a data-efficient and predictive framework for rational CS NP formulation.
- * This strategy supports Quality-by-Design (QbD) principles for pharmaceutical optimization of nanocarriers.
- * The developed methodology can accelerate the optimization process for drug delivery systems utilizing CS NPs.
