GREEN RECOVERY OF PHENOLICS FROM LUCUMA (Pouteria lucuma) SEED BY-PRODUCT USING LACTIC ACID-BASED NADES AND ADVANCED EXTRACTION TECHNOLOGIES (MAE/UPAE): A COMPARATIVE AI AND RSM MODELING STUDY

Under a circular bioeconomy framework, the upcycling of agro-industrial by-products represents a sustainable strategy for recovering high-value compounds. Lucuma (Pouteria lucuma), an Andean fruit valued for its functional properties, generates significant seed by-products during processing. Although these seeds represent an unexploited secondary raw material containing bioactive metabolites with reported antioxidant and gastroprotective potential, their sustainable valorization requires green and efficient extraction strategies. This study aimed to develop and optimize a green extraction process for bioactive compounds from lucuma seeds using natural deep eutectic solvents (NADES) through a two-stage approach involving solvent selection and process optimization.

First, four lactic acid-based NADES containing different hydrogen bond acceptors were evaluated for total polyphenol content (TPC) and DPPH antioxidant activity. Principal component analysis (PCA) explained 92.1% of the total variability and revealed a strong association supporting the selection of lactic acid–ammonium acetate (LAAA). Among five extraction strategies—microwave-assisted extraction (MAE), ultrasound bath-assisted extraction (UBAE), ultrasound probe-assisted extraction (UPAE), agitation-assisted extraction (AAE), and agitation-assisted extraction with 70% methanol (AAEOH, control)—LAAA consistently achieved the highest initial TPC and DPPH values (3.97 g GAE/100 g d.w. and 5.25 mmol TE/100 g d.w., respectively). Its physicochemical profile (density: 0.93 g/cm³, viscosity: 7.94 cP, conductivity: 6.43 mS/cm, pH: 4.19) further supported its selection for process optimization.

Significant factors (p < 0.05) for each extraction technology were incorporated into Box–Behnken designs, and simultaneous TPC and DPPH optimization was performed using response surface methodology (RSM) and artificial neural networks coupled with genetic algorithms (ANN-GA). MAE with LAAA yielded the highest experimental values (8.34 ± 0.15 g GAE/100 g d.w. for TPC; 10.47 ± 0.29 mmol TE/100 g d.w. for DPPH) in an optimized extraction time of 19.5 s. UPAE achieved 5.53 ± 0.18 g GAE/100 g d.w. and 4.91 ± 0.09 mmol TE/100 g d.w. in 40.7 min, whereas AAE yielded 3.52 ± 0.23 g GAE/100 g d.w. and 4.69 ± 0.22 mmol TE/100 g d.w. in 100 min. Relative to AAE, MAE increased TPC 2.37-fold and DPPH 2.23-fold, demonstrating that the choice of technology substantially influenced phenolic recovery. Relative to the control (AAEOH), MAE achieved 2.18-fold higher TPC and 3.91-fold higher antioxidant capacity. RSM models exhibited R² values of 82.36–96.36%, whereas ANN-GA models reached 85.53–97.16% with lower validation errors (0.48–2.63% vs. 3.85–7.81%), highlighting the value of ANN-GA for multi-response optimization.

Overall, combining LAAA with MAE maximized phenolic recovery and antioxidant activity while substantially reducing extraction time. The complementary advantages of high extraction performance and improved predictive accuracy provide a strong basis for selecting MAE and ANN-GA for valorizing lucuma seed residues.

Acknowledgments: PROCIENCIA/CONCYTEC, Contrato Nro. PE501084628-2023-PROCIENCIA.