Response Surface Methodology-Based Optimization and Simulation of Bioactive Compound Extraction from Citrullus lanatus Rind

The rind of Citrullus lanatus (watermelon), often regarded as agricultural waste, is a rich reservoir of bioactive compounds with notable antioxidant properties, making it a promising resource for applications in the cosmetic, food, and pharmaceutical industries. This study employed Response Surface Methodology (RSM) to optimize extraction conditions for maximizing the yield of phytochemicals from watermelon rind. A Central Composite Design (CCD) was utilized to evaluate the effects of key extraction parameters specifically, extraction time and temperature on total phenolic and flavonoid contents. In earlier research, four regression models Linear, Two-Factor Interaction (2FI), Quadratic (QM), and Cubic were applied to determine optimal extraction yields at low, medium, and high levels of the variables. Expanding on this, the current study introduced an additional critical factor and extended the experimental levels to five (-α, -1, 0, 1, +α), allowing for improved model fitting and more precise optimization of extraction conditions. An adaptive regression model was also incorporated to enhance the model's goodness-of-fit, reduces deviation from the zero-residual line, and improve the overall optimization outcomes. Comparative analysis, including simulations with three-dimensional surface plots, revealed that the local linear regression model 𝐿𝐿𝑅஺஻ with variable bandwidths performed comparably to the quadratic model and outperformed the other evaluated models when applied to watermelon rind as a natural coagulant. The regression and optimization results demonstrated that the proposed model slightly outperformed existing approaches under optimized conditions. Furthermore, simulation analysis was conducted to assess the stability of the regression models under a misspecification parameter 𝛾, constrained within the interval0 ≤ 𝛾 ≤ 1. The results confirmed that both QM and 𝐿𝐿𝑅஺஻ simulated models remained stable as γ increased from 0 to 1, validating the robustness of the models. Overall, the findings support the potential of watermelon rind as a sustainable and valuable source of natural antioxidants, reinforcing its applicability in food, pharmaceutical, and cosmetic product development. Keywords: Citrullus lanatus rind, Local linear regression, Locally adaptive bandwidths, Response surface methodology, Simulation.