ComDim for multi-block data analysis of Brazilian coffees: Identification of metabolites and application to the interpretation of the near infrared and 1H NMR spectral data, quality parameters and sensory properties

Common Dimension (ComDim) chemometrics method for multi-block data analysis was employed to evaluate quality parameters (granulometry, moisture, infusion time and color L* a* and b*), sensory properties (powder fragrance, drink aroma, acidity, bitterness, flavor, body, astringency, residual flavor, and overall quality), near infrared (NIR) and 1H NMR spectral data of forty eight Brazilian coffee samples divided into four groups with different qualities (Hard/Rioysh - medium roasting [1], Hard/Rioysh - very dark roasting [2], Rioysh/Rio - dark roasting [3] and Hard type 5/6-17/18/Robusta type 5/6 - light roasting [4]). ComDim was applied in the 04 data blocks corresponding to the spectral techniques and the methods of analysis of quality parameters and sensorial properties. One common dimension was sufficient to take into account all sources of variation in the four data blocks and predominant information was found in the NIR spectra and 1H NMR. ComDim results indicated that sensory properties, with body exception, are correlated with metabolites such as trigonelline, formate, caffeine, chlorogenic acids, choline, lactate, citrate, lipids, color (L* a* b*) and NIR region between 1400 to 1650 nm. In contrast, body sensory property, is correlated with metabolites such as N-methylpyridinium, quinic acids, myo-inositol, and lipids, moisture content and NIR region between 900 to 1400 nm. NIR regions corroborates with compounds identified by 1H NMR. The roasting degree was the main responsible for the discrimination of the samples. Samples with less intense roasting (1 and 4) were discriminated from the samples with more intense roasting (2 and 3). Based on the 1H NMR profiles was possible to conclude that the water extracts from coffees with different qualities were very similar in chemical composition. The proposed methodology demonstrates the applicability of ComDim for the characterization of samples when traditional methods and advanced analytical methods are used to describe the same samples.