Innovative chemometrics-based analytical approaches for the multiplatform evaluation of Brazilian Canephora coffees with geographical indication

Abstract

This project refers to a research internship abroad related to a PhD thesis in Brazil researching high-quality Brazilian Canephora coffees (Conilon and Robusta) using spectroscopy-based techniques coupled with chemometric tools for their characterization, discrimination, authentication, and quality control. For the project, Robusta Amazônico coffees from Matas de Rondônia (North region) and Conilon from Espírito Santo (Southeast region) registered with geographical indication (GI) were collected. In addition, Conilon coffees from Bahia (Northeast region) belonging to high or intermediate quality categories, and currently without GI, as well as specialty Arabica coffees of different qualities and origins were collected in order to make a comparative study and evaluate the performance of different spectroscopy techniques in discriminating them. The samples have been evaluated by different analytical techniques - almost all spectroscopic for fingerprinting: benchtop near-infrared (NIR), portable NIR, mid-Fourier transform infrared spectroscopy with attenuated total reflection (MID-ATR-FTIR), ultraviolet-visible (UV-Vis), and proton nuclear magnetic resonance (1H NMR). The exceptions are UHPLC-ESI-QTOF-MS/MS that has been used to characterize each sample with respect to its organic/metabolite composition, flame atomic absorption spectrometry (FAAS) determination of essential minerals to obtain complementary information about the inorganic composition of coffee samples and finally smartphone-based image that has been obtained directly from the roasted and ground coffees to extract color histograms for origin inspection and fraud screening of coffee. The purpose of the internship abroad is to apply innovative chemometrics-based analytical approaches to the evaluation of the data obtained from these multiple analyses, exploring novel methods in the field of chemometrics. Different supervised multi-block (data fusion) strategies will be tested with particular focus on the methods which can provide a better insight into common and distinctive portions of information in the blocks. In general, they could be used to try to discriminate the samples originating from the main producing regions of Brazil according to different quality factors, such as geographical origin, varietal, quality, GI versus non-GI Canephora or species when considering Arabica samples. Moreover, methods to analyze multivariate data from design experiments will be tested to disentangle the effect of the different sources of variation on the experimental data.