A Novel pH-Responsive Starch-Modified Polyacrylamide Hydrogel Nanocomposite reinforced with γ-Al₂O₃ for Oxaliplatin Delivery with Enhanced Anticancer Activity, pH-sensitive controlled Release, and Hybrid Kinetic–Machine Learning Modeling

• A carbohydrate-modified PAAm/Str hydrogel nanocomposite reinforced with γ-Al₂O₃ was developed for Ox delivery. • A green W/O/W double emulsion method was employed to improve encapsulation and control drug release. • The nanocarrier exhibited pH-responsive behavior with sustained release under physiological conditions and faster release in acidic environments. • Incorporation of γ-Al₂O₃ significantly enhanced drug loading and encapsulation efficiency. • Standalone ML models showed unstable predictive performance under time-aware validation. • A hybrid kinetic–ML framework improved prediction stability while preserving physical interpretability. This study developed a carbohydrate-modified polyacrylamide/starch (PAAm/Str) hydrogel nanocomposite reinforced with gamma-alumina (γ-Al₂O₃) as a delivery platform for oxaliplatin (Ox). The novelty of this work lies in integrating a double emulsion-based encapsulation strategy within a new PAAm/Str/γ-Al₂O₃ hydrogel nanocomposite, combined with a hybrid predictive framework linking classical kinetic modeling with machine learning (ML)-based residual correction. Ox was encapsulated using a green water-in-oil-in-water (W/O/W) double emulsion method, forming stable PAAm/Str/γ-Al₂O₃@Ox nanoemulsions within a dense hydrogel nanocomposite network. Structural and morphological analyses confirmed successful incorporation, spherical morphology, and high colloidal stability. The system achieved a drug loading efficiency of 42% and an encapsulation efficiency of 74%. In vitro release studies under physiological (pH 7.4) and cancer-mimicking (pH 5.4) conditions demonstrated a distinct pH-responsive release profile, with sustained release under normal conditions and accelerated release in acidic environments. To address the limitations of data-limited experimental systems, a hybrid modeling approach was employed. While standalone ML models showed unstable forward prediction under time-aware validation, the proposed framework improved predictive stability under physiological conditions by combining a Weibull kinetic model with ML-based residual learning while preserving physical interpretability. Cytotoxicity studies confirmed enhanced anticancer activity in A549 cells (lung cancer) with reduced toxicity toward L929 cells. Overall, the proposed system demonstrates controlled release, improved therapeutic selectivity, and a more reliable predictive framework for drug release behavior.