Application of GC×GC-MS in VOC analysis of fermented beverages

GC×GC is an efficient tool for the analysis of volatile compound. However, improvements are still required on VOC extraction, GC×GC setup and data processing. Different sample preparation techniques and GC×GC setup were compared based on the literature study and experimental results. Each VOC extraction technology has its own drawbacks and needs new developments. There wasn’t an ideal sample preparation technique to recover all the VOCs from the beverage sample. Furthermore, the VOCs recovered by different techniques were very different. The discussion of the pros and cons of the different techniques in our study can serve as a guide for the further development and improvement of these techniques. Combining the results from different sample preparation techniques is necessary to achieve a higher coverage of global VOC profiling. For the known fermentative aromatic compounds, the best coverage can be reached by using SPME together with SPE for beer, and VALLME for wine and cider. A fine GC×GC method development involves modulator selection, column combination and parameter optimization. Thermal modulator provides high detection sensitivity and allow exceptional trace analysis. Since the analytes coverage is the most important factor of in beverage VOC profiling, thermal modulation is a better choice. In fermented beverages, there are more polar compounds than non-polar compounds. The most suitable column combination is polar-semipolar. Same column diameters shall be used to minimize the column overloading. GC×GC parameters must be optimized. These parameters interact with each other therefore statistical prediction model is required. Response surface model is capable of doing this job while using a small number of experimental tests. The nearest neighbor distance was a suitable measurement for peak dispersion. Column and detector saturations are unavoidable if the metabolic sample is measured at one dilution level, incorrect peak deconvolution and mass spectrum construction may happen. Data processing results can be improved by a two-stage data processing strategy that will incorporate a targeted data processing and cleaning approach upstream of the “standard” untargeted analysis. Our experiments show a significant improvement in annotation and quantification results for targeted compounds causing instrumental saturation. After subtracting the saturate signal of targeted compounds, the MS construction was improved for co-eluted compounds. Incomplete signal subtraction may occur. It leads to the detection of false positive peaks or to interferences with the construction of mass spectra of co-diluted peaks. Highresolution MS libraries and more accurate peak area detection methods should be tested for further improvement.