RESUMO
An untargeted metabolomics approach based on HRMS has been applied to Colombian green coffee to develop a discrimination model to highlight the most differential compounds. For this purpose, 41 green coffee samples of different genotypes collected from 5 regions were analysed. Samples were extracted with aqueous and organic solvents to cover a wide range of compounds. Sample extracts were randomly injected and data were pre-processed with XCMS software. PCA was used to verify quality control samples behaviour, and PLS-DA and DD-SIMCA were employed to create models for discrimination using VIP variable selection method. Thirteen different compounds correctly separate green coffee samples according to their origin, several related to the quality and health benefits of coffee. Model validation was achieved using both cross-validation and an additional set with coffee samples from different harvest year. The results reveal that UHPLC-(Q)ToF MS-based metabolomics is a suitable tool to develop food origin discrimination strategies.