Optimization of extraction in supercritical fluids in obtaining seed oil by response surface methodology and artificial neuronal network coupled with a genetic algorithm.

When processing lucuma (), waste such as shells and seeds is generated, which is a source of bioactive compounds. Recently, lucuma seed (LS), especially its oily fraction, has been studied for containing phytosterols and tocopherols, powerful antioxidants with health benefits. This study proposes lucuma seed oil (LSO) extraction using supercritical fluid (SCF) to improve the quality of the extract and minimize the environmental impact. LS was previously characterized, and the extraction parameters were optimized using a Box-Behnken design, considering temperature (40-60°C), pressure (100-300 bar), and CO flow rate (3-7 mL/min), applying the response surface methodology (RSM) and neural networks with genetic algorithm (ANN+GA). The optimal parameters were 45°C, 300 bar, and 6 mL/min, obtaining 97.35% of the total oil content. The RSM and ANN+GA models showed R values of 0.9891 and 0.9999 respectively, indicating that both models exhibited a good fit to the experimental data. However, ANN+GA provided a greater proportion of the total variability, which facilitates the identification of the optimal parameters for the extraction of oil from lucuma seeds. Compared to the Soxhlet method, the LSO obtained by SCF presented better acidity (4.127 mg KOH/g), iodine (100.294 g I/100 g), and refraction indices (1.4710), as well as to a higher content of mono- and polyunsaturated fatty acids. Supercritical CO extraction is presented as a sustainable green alternative to Soxhlet extraction for extracting oil from lucuma seed due to its high extraction efficiency and similar fatty acid profile.