A Closer Look at the Bigger Picture

The latest high-throughput and powerful chromatographs and mass analyzers have found their rightful place in analytical laboratories worldwide. These instruments provide an extraordinary amount of high-quality data, especially comprehensive two-dimensional chromatography (for example, GC×GC or LC×LC) and/or high-resolution mass spectrometers.

One rising star in analytical chemistry is the combination of the pixel-based approach with non-targeted analysis. Here, we aim to look at the whole spectrum of chemical compounds in the samples by using the entire chromatogram in the data analysis, instead of individual peak picking and integrations. The great advantage is that the chemical differences between samples is highlighted in a broader way, particularly useful in petroleum and oil characterization, omics-based studies and the analysis of environmental pollution.

Our team uses a pixel-based approach to carry out oil fingerprinting analysis and to investigate environmental pollution in urban areas. We have two main goals: first, to compare samples to discover the chemical patterns that describe them and second, to simplify the extraction of the information that explains these patterns. 

As an example, non-targeted analysis has allowed us to identify a much wider range of potential anthropogenic pollutants released to the environment compared with traditional targeted analysis of conventional persistent organic pollutants (POPs). More specifically, we performed pixel-based analysis, where the main chemical variability of the samples is assessed pixel-by-pixel from entire chromatograms. There is no peak picking or peak integration steps in this approach – but the data analysis is nevertheless more complex because of the high number of chromatographic pixels.

We also use pixel-based analysis in forensic investigations of oil spills. The aim is to find the source(s) of the spill (1)(2), with positive matches of spill–source pairs obtained after comparing their respective chemical fingerprints. We use GC-MS and GC×GC-MS to analyze hydrocarbons in oil spills, and pixel-based analysis to match spills to their source (3)(4).

Pixel-based analysis is usually divided into two parts: i) preprocessing, and ii) modeling. In the first step, it is recommended that the sample chromatograms are “corrected” to minimize the influence of experimental noise. A key preprocessing step to make the pixels adequately comparable is the correction of retention time shifts that can occur because of multiple injections and column aging. The modeling step focuses the analysis on the most relevant chemical information from the chromatographic signals.

In our research, principal component analysis (PCA) has been the preferred solution when using models for pixel-based analysis, because it efficiently extracts the systematic patterns in the pixels that allow us to group the samples according to their chemical similarity. Similar samples should have similar pixel-by-pixel patterns, and will therefore belong to the same group in the principal components score plot. As an example, we investigated the entire chemical fingerprint of sediment samples analyzed by GC×GC–(HR)MS from two polluted sites (Utterslev Mose and Fortress channel) in Copenhagen, Denmark. The chemical similarity of the samples was assessed through the score plot of the PCA model. The samples were clearly grouped according to their origins in the PC1 versus PC2 subspace, which means the chemical fingerprints were different in each place.

We therefore encourage the use of pixel-based analysis for a deeper understanding of the complex chemical fingerprints analyzed by today’s high-end instruments, so that we get a closer look at the bigger “chemical” picture we have in our hands in non-targeted analysis.