Abstract
Coffee processing wastes, including defected green beans (GB) and spent coffee grounds (SCG), are underutilized by-products rich in bioactive compounds with promising applications in nutraceuticals and cosmetics. This study profiled metabolites from Arabica and Robusta GB and SCG using five solvents of varying polarity (ethanol, ethyl acetate, toluene, xylene, and hexane). Quantification of chlorogenic acid and caffeine was performed using HPLC-DAD, while comprehensive metabolite profiling was conducted via UHPLC-HRMS/MS and GC-MS. Principal Coordinate Analysis (PCoA) based on Bray-Curtis distance was applied to provide an unsupervised overview of variation according to solvent polarity and raw material type, while a supervised Random Forest (RF) model was used to assess classification performance and group-level consistency. Non-polar solvents tended to recover fatty acids and sterols, especially from SCG, whereas polar solvents such as ethanol extracted higher amounts of hydrophilic antioxidants, including chlorogenic acid from GB. Molecular networking (MN) visualized structurally related metabolite clusters and illustrated solvent- and material-associated distributions. Overall, these findings indicate that extraction method and raw material origin shape metabolite diversity and functional potential in coffee waste materials. The combined use of MN, multivariate statistics, and machine learning offers a complementary strategy for chemical mapping and provides a framework to guide future valorization of coffee by-products, while additional bioactivity validation will be required to establish specific applications.