Brazil is the largest producer, exporter and the second largest consumer of coffee in the world. The two most economically valuable species are Coffea arabica and Coffea canephora, which differ in their physical, chemical and, consequently, market value, quality and consumer acceptance. The objective of the present study was to differentiate arabica and conilon coffee species from the physicochemical properties of moisture, pH, titratable total acidity, soluble solids, total sugars, reducing and non-reducing sugars, total phenolics, chlorogenic acid, caffeine and trigonelline using Partial Least Squares Discriminant Analysis (PLS-DA). The study of the physicochemical analyzes of both species allows a greater knowledge of the constituents that discriminate them, being able to help in the identification of compounds capable of influencing the quality of the beverage, enhancing what each species has to offer the best in the elaboration of coffee blends arabica and conilon, besides the quantification of compounds that, because they are distinct in arabica coffee and conilon, can also be used in the identification of frauds. The model obtained by the PLS-DA showed a good separation between classes, that is, it effectively classified the arabica and conilon coffee samples, resulting in obtaining values of accuracy and accuracy of the model above 90 %. The variables that had the greatest weight, according to PLS-DA, in the discrimination of coffee species were pH, acidity, total phenolic compounds and caffeine.
