In response to the need from the food industry for new analytical solutions, a fit-for-purpose quantitative 1H NMR methodology was developed to authenticate pure coffee (100% arabica or robusta) as well as predict the percentage of robusta in blends through the study of 292 roasted coffee samples in triplicate. Methanol was chosen as the extraction solvent, which led to the quantitation of 12 coffee constituents: caffeine, trigonelline, 3- and 5-caffeoylquinic acid, lipids, cafestol, nicotinic acid, N-methylpyridinium, formic acid, acetic acid, kahweol, and 16-O-methylcafestol. To overcome the chemical complexity of the methanolic extract, quantitative analysis was performed using a combination of traditional integration and spectral deconvolution methods. As a result, the proposed methodology provides a systematic methodology and a linear regression model to support the classification of known and unknown roasted coffees and their blends.
Keywords: adulteration; arabica; blends; coffee; quantitative NMR; robusta.