Application of Correlation Analysis to Classify Ground Coffee Bean Extracts Using UPLC-HRMS in MSE Mode

Significance of the Topic

Coffee is a highly complex beverage whose flavor and aroma derive from hundreds of chemical constituents. Reliable classification of Arabica and Robusta beans, as well as assessment of roasting level, is critical for quality control and detection of adulteration in the global coffee supply chain.

Objectives and Study Overview

This study applied ultraperformance liquid chromatography coupled to high-resolution mass spectrometry (UPLC-HRMS) in MSE mode to blind ground coffee bean extracts. The goals were to identify discriminating metabolites, classify samples by species (Coffea arabica vs. Coffea canephora) and infer roasting intensity.

Methodology and Instrumentation

Samples A–D were brewed, centrifuged, and injected (1 µL) under positive and negative electrospray as well as APCI conditions. Chromatographic separation used an ACQUITY UPLC HSS T3 column (1.8 µm, 2.1×150 mm) over a 16 min gradient (water/acetonitrile with 0.1% formic acid). Mass spectra were acquired on a Xevo G2-XS QToF (50–1,550 m/z) using low-energy (4 eV) and high-energy ramps (10–45 eV). Data processing employed Progenesis QI with correlation analysis to group compounds by abundance patterns.

Main Results and Discussion

Reproducible UPLC-MS profiles revealed over 300 annotated compounds (mass error <5 ppm, isotope match >85%, fragment score >10). Correlation analysis identified nine clusters of metabolites showing distinct abundance trends across samples. Cluster 6 contained Mozambioside, a bitter furokaurane glycoside specific to Arabica, detected at high levels in samples A and B, confirming their Arabica origin. Clusters of dicaffeoylquinic acid isomers (DiCQAs) showed low abundance in sample D, indicating extensive roasting and conversion to chlorogenic acid lactones. Further, the ratio of trigonelline to nicotinic acid distinguished roasting between Robusta samples C and D—sample C retained high trigonelline, suggesting minimal roast, while sample D accumulated nicotinic acid.

Practical Benefits and Applications

• Rapid, multiplexed profiling of coffee extracts without targeted assays.
• Accurate classification of bean species for authenticity testing.
• Insights into roasting degree via metabolite transformations.
• Potential for routine QA/QC in food and beverage industries.

Future Trends and Opportunities

Building on this workflow, future developments may include integration of machine learning for automated classification, expansion into wider food metabolomics applications, and real-time monitoring of processing parameters. Coupling with larger reference databases and cloud-based analytics could further enhance throughput and robustness.

Conclusion

This study demonstrates that UPLC-HRMS in MSE mode combined with correlation analysis provides a powerful strategy for distinguishing coffee bean species and assessing roasting levels. The approach is rapid, reproducible, and applicable to authenticity testing and quality control in the food sector.

Instrumentation Used

• ACQUITY UPLC I-Class PLUS System
• ACQUITY UPLC HSS T3 column (1.8 µm, 2.1×150 mm)
• Xevo G2-XS QToF Mass Spectrometer
• MassLynx MS Software
• Progenesis QI Software (v2.3)

References

1. Lang A. et al. J. Agric. Food Chem. 63 (2015): 10492–10499.
2. Farah A. et al. J. Agric. Food Chem. 53 (2005): 1505–1513.
3. Lima T. C. et al. J. Agric. Food Chem. 64 (2016): 2361–2370.
4. Trugo L. C. In Caballero B., Trugo L. C., Finglas P. M. (Eds.), Encyclopedia of Food Sciences and Nutrition, Academic Press, Oxford, 2003, Vol. 7:498.
5. Stennert A., Maier H. G. Z. Lebensm. Unters. Forsch. 1994, 199:198.