From Agro-Food Waste to Functional Ingredients: A Computational Framework for Green Solvent Selection and Bioactive Compound Recovery

Agro-food residues contain valuable bioactives, but solvent selection requires integrating solubility, hazard information, and intended applications. This study developed a traceable computational framework for solvent screening and experimental prioritization. A library of 17 bioactives associated with five residue matrices and 10 solvents defined 170 potential compound–solvent pairs. The curated database comprised 911 published records, supporting reference-temperature comparisons for 37 pairs at 298.15 K. Three temperature correlations were evaluated through within-source validation. System-level comparisons detected no differences after multiplicity adjustment, whereas exploratory aggregation across eight studies favored the quadratic model over modified Apelblat for overall and endpoint prediction errors. Limited thermal coverage and dependence between datasets constrain these findings. Integrating solubility with separate CHEM21 indicators identified ten dominated options whose classifications persisted under temperature-removal sensitivity analysis. Water–ethanol comparisons demonstrated that composition choices depend on the concentration basis. A caffeine scenario connected published coffee-residue loading with ideal solvent-capacity requirements. Ambiguous chlorogenic-acid mixtures were withheld from principal composition analyses. The framework identifies conditional candidates and evidence gaps rather than globally optimal solvents. It supports targeted recovery studies while requiring independent measurements, matrix-specific extraction experiments, and application-specific regulatory assessment before conclusions about extraction performance, sustainability, or functional-ingredient efficacy can be established under practical processing conditions.