Advances in Instrumentation and Chemometrics of Two-Dimensional Gas Chromatography Systems

Abstract

Comprehensive two-dimensional (2D) gas chromatography (GC×GC) has gained significant popularity as an analysis tool for complex samples in metabolomics, petroleum science, and other fields. GC×GC has been shown to perform better than one-dimensional GC, including a 10-fold increase in peak capacity and more chemical selectivity. The work presented herein focuses on two aspects of GC×GC systems: advancements in instrumentation and chemometrics. While GC×GC is used to analyze volatile compounds, when it comes to very light analytes, like permanent gasses, new instrumental designs are needed, as widely applied wall coated open tubular (WCOT) columns do not work well. Also, thermal modulators, which are in most commercially available instruments, cannot trap such low boiling point analytes due to temperature limitations. Porous layer open tubular (PLOT) columns and flow modulators must be used for such samples. However, PLOT columns are not widely used in GC×GC due to their extreme retention on analytes, and hence being hard to pair with WCOT columns due to temperature differences that are needed for analytes to elute. In this work, we present a new GC×GC system with a PLOT column in the first dimension and a WCOT column in the second dimension. This system also used a high temperature diaphragm valve as a modulator. We showed the ability of this system to separate heavy mixtures like gasoline, and by trying thinner film PLOT columns, we showed the potential of this system to be used with even higher boiling point analytes. GC×GC is an essential analytical technique for biological samples, as they are complex and make a perfect sample for this technique. However, due to the biological variation that is possible in these samples, data analysis can be tricky. Here, we present three projects that analyzed biological samples: cycling yeast, moisture-damaged beans, and VOCs of Malassezia pachedermatis. While all were very different, they were analyzed using the tile-based Fisher ratio and then post-processed to answer specific biological questions of the dataset. When analyzing cycling yeast, we were looking for analytes that were not only changing between different classes but also had a specific cycling pattern. This was achieved by simulating random data to analyze which analytes were like random patterns and which had some underlying biological information. The second dataset presented is moisture-damaged cacao beans, where the F-ratio method had to be altered, so not only the analytes that were changing with the molding process would be found, but also the hits that were bean origin-specific and did not change with molding. Lastly, we analyzed VOCs produced by M. pachedermatis, when grown at three different pH. In this case, the analytes need to be classified into different types, based on how much and how they changed in signal between blank media and M. pachedermatis, and to be analyzed for potential pH-dependency. This was performed by post-processing using R-metric and RSD-metric.