Finding needles in haystacks: Using statistical tools for complex samples measured by using GCxGC-HRMS

Finding needles in haystacks: Using statistical tools for complex samples measured by using GCxGC-HRMS

A. John Dane* 1; Scott J. Campbell 2; John Moncur 2; Kirk R Jensen 1; Robert B Cody 1

1 JEOL USA, Inc., Peabody, MA, USA; 2 SpectralWorks Ltd, Runcorn, United Kingdom

.First published ASMS, June 2024.

Introduction

Complex materials like petroleum samples can be difficult to analyze by traditional gas chromatography-mass spectrometry (GC-MS). Furthermore, given the complexity of these materials which can contain hundreds or even thousands of compounds, it can be difficult to determine differences between samples when measured using traditional GC-MS methods.

To address this problem, two-dimensional gas chromatography (GCxGC) combined with high-resolution time-of-flight mass spectrometry (HRTOFMS) can be used to provide a second dimension of separation for compounds that coelute using 1D GC as well as high mass accuracy to provide molecular formula information for each analyte.  Consequently, using these techniques together is an effective way to measure highly complex samples like petrochemical products.

In this study, four different diesel samples were measured by using GCxGC-HRTOFMS and were then directly compared using new software capabilities involving new statistical data analysis strategies to identify unique compounds in the diesel samples.

Table 1. Experimental Parameters

Figure 1. AccuTOF GC-Alpha Resolving Power: 30,000
Mass Accuracy: ≤ 1ppm
Figure 2. NEW SepSolve INSIGHT-Thermal fully programmable GCxGC thermal modulation system.
Figure 3. Representative GCxGC-EI TICC contour plots for each diesel sample measured for this study.

Discussion

Figure 3 shows TICC contour plots for each diesel brand measured by using GCxGC-EI. The SpectralWorks AnalyzerPro XD software was then used to look at the data, and produced a PCA plot comparing all four samples to each other (Figure 4).  These results highlighted the fact that the samples were different from each other despite the fact that their GCxGC contour plots showed many common features.

Additionally, volcano plots (Figure 5 and 6) were used to identify unique compounds in each sample when compared individually to each other. Analytes demarked by green triangles are unique to the sample on the left and red triangles are unique to the sample on the right.

Figure 4. PCA plot for the 4 different diesel brands tested in this study show a clear separation of the 4 different diesel
Figure 5. The volcano plot shows the comparison of 2 diesel brands. Additionally, the charts on the right show that the selected compound is also unique to the first diesel sample amongst the 4 samples.
Figure 6. The volcano plot shows the comparison of 2 diesel brands. Additionally, the charts on the right show that the selected compound is not present in only one of the samples amongst the 4 samples.

Conclusions

• PCA was able to separate each brand into tight clusters, confirming that there are unique features in each sample.

• Despite the complexity of the diesel samples (over 850 compounds detected), there were compounds that were uniquely present (and uniquely absent) that the data analysis software was able to quickly identify.