Ph-Triggered Curcumin Delivery Via a Hybrid Starch–Gqd–Γ-Al₂o₃ Nanocarrier: A Machine Learning Study Publisher



Khajeh MM ; Pourmadadi M ; Abad FRA ; Najafi M ; Adeli F ; Mehrabi MG ; Rahdar A ; Abdouss M ; Fathikarkan S ; Pandey S
Source: Journal of Cluster Science Published:2026

Abstract

The clinical utility of curcumin in oncology is hindered by its poor aqueous solubility, low bioavailability, and rapid systemic clearance. In this study, a pH-responsive hybrid hydrogel nanocarrier composed of starch, graphene quantum dots (GQD), and gamma(γ)-alumina (γ-Al₂O₃) was engineered to improve the delivery, stability, and tumor selectivity of curcumin. GQD and γ-Al₂O₃ were synthesized and incorporated into a starch-based hydrogel, followed by curcumin loading. The nanocarrier was comprehensively characterized using Fourier-transform infrared spectroscopy (FTIR), X-ray diffraction (XRD), field-emission scanning electron microscopy (FE-SEM), dynamic light scattering (DLS), and zeta potential analysis. Drug loading capacity, encapsulation efficiency, pH-dependent release kinetics at acidic and physiological pH values, and in vitro cytotoxicity against U-87 MG glioblastoma cells and L929 fibroblasts were evaluated. The optimized nanocarrier exhibited nanoscale dimensions between 147 and 193 nm, high colloidal stability with a zeta potential of + 49 mV, a loading capacity of 46%, and an encapsulation efficiency of 86%. Curcumin release was markedly pH-dependent, with 94% release under acidic conditions and 38% release at physiological pH over 96 h. Kinetic analysis indicated anomalous transport in acidic conditions and zero-order release behavior at physiological pH. Cytotoxicity assays revealed pronounced anticancer activity with minimal toxicity toward normal fibroblasts. In addition, experimental release data were used to train and test multiple machine learning models, including decision trees, linear and polynomial regression, random forest, support vector machines, k-nearest neighbors, and neural networks. Model performance was evaluated using the coefficient of determination and mean absolute error. Tree-based ensemble models and neural networks demonstrated superior predictive accuracy compared to classical kinetic models. These findings suggest that the Starch–GQD–γ-Al₂O₃ hybrid hydrogel is a promising platform for pH-responsive, targeted curcumin delivery and demonstrate the utility of machine learning as a predictive tool for optimizing drug delivery systems. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2026.