Integration of COSMO-RS and ANN-GA for extraction of echinacoside and acteoside from Cistanche deserticola using natural deep eutectic solvents.

A novel method for green extraction of echinacoside and acteoside from Cistanche deserticola (C. deserticola) using natural deep eutectic solvents (NADES) has been developed, integrating conductor-like screening model for realistic solvents (COSMO-RS) and artificial neural network-genetic algorithm (ANN-GA) optimization to maximize efficiency. Among 48 NADES evaluated, choline chloride-propylene glycol (1:3 M ratio, 30 wt% water content) was identified as the optimal solvent through COSMO-RS and single-factor optimization. The ANN model (3-9-1 topology, R > 0.99) efficiently captured nonlinear relationships between extraction parameters, with GA identifying the optimal conditions. The maximum yields of echinacoside and acteoside under optimized conditions were 7.84 ± 0.28 mg/g and 1.12 ± 0.02 mg/g, respectively, which were 1.97-3.98 times higher than those obtained with conventional solvents. This work demonstrates the synergistic potential of COSMO-RS and ANN-GA for solvent screening and extraction process optimization, offering a scalable framework for extracting other bioactive substances.