Chemometric Authentication of Brazilian Coffees Based on Chemical Profiling

Targeted and Non-Targeted HPLC Analysis of Coffee-Based Products as Effective Tools for Evaluating the Coffee Authenticity

Coffee is a very popular beverage worldwide. However, its composition and characteristics are affected by a number of factors, such as geographical and botanical origin, harvesting and roasting conditions, and brewing method used. As coffee consumption rises, the demands on its high quality and authenticity naturally grows as well. Unfortunately, at the same time, various tricks of coffee adulteration occur more frequently, with the intention of quick economic profit. Many analytical methods have already been developed to verify the coffee authenticity, in which the high-performance liquid chromatography (HPLC) plays a crucial role, especially thanks to its high selectivity and sensitivity. Thus, this review summarizes the results of targeted and non-targeted HPLC analysis of coffee-based products over the last 10 years as an effective tool for determining coffee composition, which can help to reveal potential forgeries and non-compliance with good manufacturing practice, and subsequently protects consumers from buying overpriced low-quality product. The advantages and drawbacks of the targeted analysis are specified and contrasted with those of the non-targeted HPLC fingerprints, which simply consider the chemical profile of the sample, regardless of the determination of individual compounds present.

Metabolomics-Based Approach for Coffee Beverage Improvement in the Context of Processing, Brewing Methods, and Quality Attributes

Coffee is a worldwide beverage of increasing consumption, owing to its unique flavor and several health benefits. Metabolites of coffee are numerous and could be classified on various bases, of which some are endogenous to coffee seeds, i.e., alkaloids, diterpenes, sugars, and amino acids, while others are generated during coffee processing, for example during roasting and brewing, such as furans, pyrazines, and melanoidins. As a beverage, it provides various distinct flavors, i.e., sourness, bitterness, and an astringent taste attributed to the presence of carboxylic acids, alkaloids, and chlorogenic acids. To resolve such a complex chemical makeup and to relate chemical composition to coffee effects, large-scale metabolomics technologies are being increasingly reported in the literature for proof of coffee quality and efficacy. This review summarizes the applications of various mass spectrometry (MS)- and nuclear magnetic resonance (NMR)-based metabolomics technologies in determining the impact of coffee breeding, origin, roasting, and brewing on coffee chemical composition, and considers this in relation to quality control (QC) determination, for example, by classifying defected and non-defected seeds or detecting the adulteration of raw materials. Resolving the coffee metabolome can aid future attempts to yield coffee seeds of desirable traits and best flavor types.

Near-Infrared Spectroscopy Applied to the Detection of Multiple Adulterants in Roasted and Ground Arabica Coffee

Roasted coffee has been the target of increasingly complex adulterations. Sensitive, non-destructive, rapid and multicomponent techniques for their detection are sought after. This work proposes the detection of several common adulterants (corn, barley, soybean, rice, coffee husks and robusta coffee) in roasted ground arabica coffee (from different geographic regions), combining near-infrared (NIR) spectroscopy and chemometrics (Principal Component Analysis—PCA). Adulterated samples were composed of one to six adulterants, ranging from 0.25 to 80% (w/w). The results showed that NIR spectroscopy was able to discriminate pure arabica coffee samples from adulterated ones (for all the concentrations tested), including robusta coffees or coffee husks, and independently of being single or multiple adulterations. The identification of the adulterant in the sample was only feasible for single or double adulterations and in concentrations ≥10%. NIR spectroscopy also showed potential for the geographical discrimination of arabica coffees (South and Central America).

Classification of organic and ordinary kiwifruit by chemometrics analysis of elemental fingerprint and stable isotopic ratios

Elemental fingerprint, stable isotopic analysis, and chemometrics were combined to identify organic kiwifruit from ordinarily cultivated kiwifruit. Samples of organic (n1 = 78) and ordinary kiwifruit (n2 = 85) were collected from neighboring areas. For elemental fingerprint, the contents of 15 elements in fresh fruits, including Al, Cr, Mg, Pb, Zn, Ca, Cu, Mn, Se, Cd, Fe, Na, Sr, Co, and K, were determined by inductively coupled plasma optical emission spectrometry (ICP-OES). Three stable isotopes, including δ¹³ C, δ¹⁵ N, and δ¹⁸ O, were analyzed using an isotope-ratio mass spectrometer (IRMS). Different classification methods including soft independent modeling of class analogy (SIMCA), partial least squares discriminant analysis (PLSDA), and least squares support vector machines (LS-SVM), were used to discriminate the organic and ordinary kiwifruits by fusion of elemental and stable isotopic. As a result, the sensitivity, specificity, and overall accuracy of SIMCA model were 0.885, 0.857, and 0.864, respectively. PLSDA and LS-SVM obtained 0.950 and 0.983 classification accuracy of organic and ordinary kiwifruits, respectively. It was demonstrated that elemental fingerprint and stable isotopic analysis would provide useful chemical information for the identification of organic fruits, and the capacity of these methods could be enhanced by chemometrics. PRACTICAL APPLICATION: The classification of kiwifruit usually relies on the label assigned by the merchant, which is prone to deceive consumers. This research has developed an accurate and effective classification method based on stable isotopes and mineral elements for the identification of ordinary kiwifruit and organic kiwifruit, providing a tool for the quality monitoring of organic food.

Use of non-volatile compounds for the classification of specialty and traditional Brazilian coffees using principal component analysis

Coffee beans contain different volatile and non-volatile compounds that are responsible for their flavor and aroma. Herein, principal component analysis (PCA) was employed to correlate the non-volatile composition of specialty and traditional coffees with drink quality. The quantified non-volatile compounds included caffeine, chlorogenic acid, caffeic acid, and nicotinic acid in both types of coffee samples, while 5-hydroxymethylfurfural was only quantified in the specialty coffee samples. The most abundant compounds present in specialty coffees were associated with the aroma and flavor, affording a high drink quality. In traditional coffees, the most abundant compounds included nicotinic acid and caffeine, indicating a stronger roasting process, loss of sensory characteristics, and blended formulations. PCA showed a distinction between the traditional and specialty coffees such that a relationship between the contents of the compounds in each type of coffee, quality, and classification could be established.

Procrustes Cross-Validation of short datasets in PCA context

We suggest using a new tool, Procrustes cross-validation, as an alternative to a regular cross-validation for short datasets where each sample is important and, therefore, cannot be removed in line with the conventional leave-one-out cross-validation procedure. The advantages of the new approach are demonstrated using two real-world examples: the first one contains discrete variables (chemical profiles). The second one is based on continuous data (spectra). The method is implemented in R and Matlab as a small procedure that any analyst can easily use.

Integrating a Low-Cost Electronic Nose and Machine Learning Modelling to Assess Coffee Aroma Profile and Intensity

Aroma is one of the main attributes that consumers consider when appreciating and selecting a coffee; hence it is considered an important quality trait. However, the most common methods to assess aroma are based on expensive equipment or human senses through sensory evaluation, which is time-consuming and requires highly trained assessors to avoid subjectivity. Therefore, this study aimed to estimate the coffee intensity and aromas using a low-cost and portable electronic nose (e-nose) and machine learning modeling. For this purpose, triplicates of nine commercial coffee samples with different intensity levels were used for this study. Two machine learning models were developed based on artificial neural networks using the data from the e-nose as inputs to (i) classify the samples into low, medium, and high-intensity (Model 1) and (ii) to predict the relative abundance of 45 different aromas (Model 2). Results showed that it is possible to estimate the intensity of coffees with high accuracy (98%; Model 1), as well as to predict the specific aromas obtaining a high correlation coefficient (R = 0.99), and no under- or over-fitting of the models were detected. The proposed contactless, nondestructive, rapid, reliable, and low-cost method showed to be effective in evaluating volatile compounds in coffee, which is a potential technique to be applied within all stages of the production process to detect any undesirable characteristics on–time and ensure high-quality products.

Analytical strategy coupled to chemometrics to differentiate Camellia sinensis tea types based on phenolic composition, alkaloids, and amino acids

Catechins, amino acids, and alkaloids are primary chemical components of tea and play a crucial role in determining tea quality. Their composition and content largely vary among different types of tea. In this study, a convenient chemical classification method was developed for six Camellia sinensis tea types (white, green, oolong, black, dark, and yellow) based on the quantification of their major components. Twenty-one free amino acids, 6 catechins, 2 alkaloids, and gallic acid in 24 teas were quantified using ultra-high-performance liquid chromatography (UHPLC). The total catechin contents in these tea samples ranged from 10.96 to 95.67 mg/g, while total free amino acid content ranged from 2.63 to 25.89 mg/g. Theanine (Thea) was the most abundant amino acid in all tea varieties. Catechin and amino acid levels in tea were markedly reduced upon fermentation of tea. Furthermore, high-temperature processing (roasting) during tea production induced degradation and epimerization of catechins, yielding epimerized catechins, simple catechins, and gallic acid. Principal component analysis revealed that major ester-catechins (EGCG and ECG), major amino acids (Thea), and major alkaloids (caffeine) are potential factors for distinguishing different types of tea. Linear discriminant analysis showed that 100% of teas were correctly classified in which (+)-catechin, ECG, EGC, gallic acid, GABA, cysteine, lysine, and threonine were the most discriminating compounds. This study shows that quantification of the major tea components combined with chemometric analysis, can serve as a simple, convenient, and reliable approach for classifying tea according to fermentation level. PRACTICAL APPLICATION: Different Camellia sinensis tea types can be produced worldwide but it is still challenging to know which chemical markers can be used to trace their production. in this paper we used a targeted methodology to classify six tea types (white, green, oolong, black, dark, and yellow) based on phenolic composition, alkaloids, and amino acids. The main chemical markers responsible for the discrimination were pinpointed with the use of chemometric tools.

Heat-induced formation of N,N-dimethylpiperidinium (mepiquat) in Arabica and Robusta coffee

N,N-dimethylpiperidinium (mepiquat) is a new process-induced compound formed from natural constituents during the cooking process. Mepiquat was first found in coffee and cereal products, but its formation mechanism in coffee is still unclear. In the current study, Arabica and Robusta coffee beans were roasted at different temperatures (215, 220, and 230 °C) to study the effect of roasting process on mepiquat formation. The highest mepiquat content, 1,020 µg/kg, was found in dark roast (230 °C) Indonesia Wahana, while 430 µg/kg of mepiquat was detected in medium roast (220 °C) Vietnam Robusta. At the same roasting temperature, higher level of mepiquat was observed in Arabica than in Robusta. In both species, substances related to mepiquat formation, including betaine, choline, trigonelline, lysine, carnitine, pipecolic acid (PipAc), pipecolic acid betaine (PipBet), were also detected. The lysine-based Maillard reaction and decarboxylation in Arabica and Robusta promoted mepiquat formation through the degradation of choline and trigonelline, and the formation of intermediate products. Results from both the model system and selected commercial beans showed that choline and trigonelline had a significant correlation (P < 0.01) with mepiquat formation in Arabica. PRACTICAL APPLICATION: Mepiquat is considered as a new process-induced compound resulting from typical roasting conditions, but its formation mechanism in coffee is still unclear. This work demonstrates the formation mechanism of mepiquat by many precursor substances contained in Arabica and Robusta. It is very important to figure out how mepiquat is ''naturally" present in daily diets, especially in those processed at high temperatures.

Reliable Discrimination of Green Coffee Beans Species: A Comparison of UV-Vis-Based Determination of Caffeine and Chlorogenic Acid with Non-Targeted Near-Infrared Spectroscopy

Species adulteration is a common problem in the coffee trade. Several attempts have been made to differentiate among species. However, finding an applicable methodology that would consider the various aspects of adulteration remains a challenge. This study investigated an ultraviolet–visible (UV-Vis) spectroscopy-based determination of caffeine and chlorogenic acid contents, as well as the applicability of non-targeted near-infrared (NIR) spectroscopy, to discriminate between green coffee beans of the Coffea arabica (Arabica) and Coffea canephora (Robusta) species from Java Island, Indonesia. The discrimination was conducted by measuring the caffeine and chlorogenic acid content in the beans using UV-Vis spectroscopy. The data related to both compounds was processed using linear discriminant analysis (LDA). Information about the diffuse reflectance (log 1/R) spectra of intact beans was determined by NIR spectroscopy and analyzed using multivariate analysis. UV-Vis spectroscopy attained an accuracy of 97% in comparison to NIR spectroscopy’s accuracy by selected wavelengths of LDA (95%). The study suggests that both methods are applicable to discriminate reliably among species.

Authentication of the Origin, Variety and Roasting Degree of Coffee Samples by Non-Targeted HPLC-UV Fingerprinting and Chemometrics. Application to the Detection and Quantitation of Adulterated Coffee Samples

In this work, non-targeted approaches relying on HPLC-UV chromatographic fingerprints were evaluated to address coffee characterization, classification, and authentication by chemometrics. In general, high-performance liquid chromatography with ultraviolet detection (HPLC-UV) fingerprints were good chemical descriptors for the classification of coffee samples by partial least squares regression-discriminant analysis (PLS-DA) according to their country of origin, even for nearby countries such as Vietnam and Cambodia. Good classification was also observed according to the coffee variety (Arabica vs. Robusta) and the coffee roasting degree. Sample classification rates higher than 89.3% and 91.7% were obtained in all the evaluated cases for the PLS-DA calibrations and predictions, respectively. Besides, the coffee adulteration studies carried out by partial least squares regression (PLSR), and based on coffees adulterated with other production regions or variety, demonstrated the good capability of the proposed methodology for the detection and quantitation of the adulterant levels down to 15%. Calibration, cross-validation, and prediction errors below 2.9%, 6.5%, and 8.9%, respectively, were obtained for most of the evaluated cases.