Optimization of Green-synthesized Graphene Quantum Dots using Machine Learning for Neuroprotective Antioxidant Applications

Oxidative stress caused by reactive oxygen species (ROS) plays a major role in neurodegenerative disorders. However, many engineered nanoparticles exhibit unpredictable cytotoxicity, which limits their sustainable biomedical applications. To address this challenge, a novel machine learning-guided nano- Quantitative Structure–Activity Relationship framework (nano-QSAR) was developed to identify low-cytotoxic nanomaterials prior to experimental synthesis, thereby reducing the experimental trials and improving sustainable nanomaterial design. A curated dataset of 443 nanoparticles was analyzed using multiple regression algorithms, where tuned XGBoost showed the best predictive performance R² = 0.78, Root Mean Square Error (RMSE = 6.82, Mean Absolute Error (MAE) = 4.95). Based on these predictions, graphene-based nanomaterials were selected for experimental validation. Subsequently, graphene quantum dots derived from Terminalia arjuna bark (GQD-TA) were synthesized through an eco-friendly hydrothermal method and optimized using central composite design, while ANOVA confirmed the statistical significance of the model F = 2.55, p = 0.0380. The optimized nanoparticles exhibited an average hydrodynamic size of 11 nm and were mainly composed of carbon (67.85 wt%), nitrogen (15.24 wt%), and oxygen (11.03 wt%). Antioxidant evaluation demonstrated concentration-dependent DPPH scavenging activity of 94–95% at 150 µg/mL. Furthermore, MTT assay using SH-SY5Y neuronal cells confirmed high cytocompatibility, maintaining 88.5% cell viability even at 500 µg/mL. The developed graphene quantum dots can be utilized in neuroprotective therapies, targeted drug delivery, bioimaging, and neuronal tissue engineering applications. Their high biocompatibility and low cytotoxicity make them promising candidates for sustainable nanomedicine applications.