In this installment of our GC×GC Detective Stories series, Jean-François Focant traces his journey from measuring trace pollutants to investigating the chemical clues in patient samples. He discusses how volatile compounds could support disease detection and treatment monitoring, why the clinical contribution of comprehensive two-dimensional gas chromatography (GC×GC) may depend on simpler instruments, and the data harmonization challenge that still needs solving.
How did you first become interested in GC×GC?
I started my PhD in 1998, working on dioxins and polychlorinated biphenyls (PCBs). When you measure dioxins and PCBs, you must work at the parts-per-trillion level, which creates a lot of sample preparation challenges because you have to concentrate the sample to access such low levels of dioxins and PCBs. I was working mainly with one-dimensional gas chromatography (GC) and mass spectrometry, using sector instruments in selected ion monitoring mode. But when I was doing my literature review at the beginning of my PhD, I came across some of the first publications on GC×GC. At that time, it was being presented as a possible alternative to simplify the cleanup procedure when working with very complex matrices.
The idea was that, because of the two dimensions, you could reduce the time spent getting rid of matrix effects, because the matrix components would be separated from the compounds of interest in the two-dimensional chromatographic plane. That's what interested me at the time.
Back in June 1999, as Belgium was hit by the dioxin crisis, I was visiting Don Patterson Jr. at the dioxin lab at the US Centers for Disease Control and Prevention (CDC) to learn about automating sample preparation to increase our lab capacity for better crisis management. I also knew that Don had been doing some early GC×GC work on dioxins and that John Dimandja, a former student of John Phillips, the inventor of GC×GC, was also at the CDC by then. I met John there, and he irrevocably contaminated me with his enthusiasm for the technique.
I asked whether I could spend some time at the CDC with him to explore the technology. He said yes, so I went there for three months. I was super excited, but we only managed to produce one chromatogram in that time because the coupling and modulation were so poor. But I went back for three more summers to better learn the technique. Then I finished my PhD and went back to the CDC for a two-year postdoc with Don Patterson, using GC×GC for human biomonitoring of multiple persistent organic pollutants.
So, to answer the question, I first heard about GC×GC as a tool to simplify the cleanup procedure, which wasn't really what it could do. If you inject dirt into your GC system, the liner is still a liner, and when it gets dirty, it gets dirty.
The real discovery for me was the ability to use the chromatographic space to look at dioxins, PCBs, pesticides, and flame retardants all in a single injection. That's really where the added value was for me.
Since then, I've been using GC×GC in the lab. And nowadays, we only do GC×GC – we don't use one-dimensional GC anymore – to investigate our separation science challenges.
How has GC×GC evolved over the past 25 years?
When I first went to the CDC, I was using the sweeper – that was the name of the modulator at the time. It was a moving-part modulator, which probably worked 10 percent of the time. The other 90 percent, you were trying to fix issues with it. Today, any modulator will do the job on a routine basis.
But the most important changes have probably been on the software side. Today, you can do data processing, data reduction, and alignment using commercial software from LECO, GC Image, or ChromSpace, for example.
In short, you can be up to speed immediately on a modern GC×GC system. It's not just a research tool anymore; it can really be used routinely – and that wasn't the case 15 years ago.
How does GC×GC compare with other analytical techniques today?
I think the evolution of GC×GC was probably slower than some other techniques. One of the main reasons was competition among the people who were there at the beginning, where there was a lot of reluctance to share information and new advances. As a young scientist, I was surprised by that because – and perhaps I'm naïve – when I do something, I want to share it with everyone because I think that's what science is about. I think that slowed down the evolution and development of the technique, and it's one of the reasons why GC×GC was seen as an exotic technology for a long time.
Thankfully, that's no longer the case. You can now find GC×GC-MS in routine industrial laboratories.
Has GC×GC reached a turning point in recent years?
There's definitely been a change. In previous years, when you went to a GC×GC meeting or listened to a GC×GC presentation at the International Symposium on Capillary Chromatography or elsewhere, people would talk about the new modulator or software they’d developed, or that they’d been trying GC×GC to solve this problem.
We’re now in a new era, which is even more exciting because we see GC×GC being used, for example, as a real metabolomics tool – not an exotic trial from an exotic lab somewhere. Some metabolomics labs really use GC×GC daily now. Indeed, we do a lot of medical applications in our lab that wouldn’t be possible without GC×GC.
But it's not that we've reached a plateau now that the hardware is quite stable and we can use it routinely. A lot of know-how and manual input are still needed to process data. It's easy to generate a chromatogram, a peak table, or an image. But if you're looking for biomarkers of exposure, for example, from the more than 1,000 features you get, you need to isolate the important molecules.
People know how to do that now, but everyone has their own recipe. If I take a dataset and give it to you, as an expert GC×GC user, you'll process it with your own methodology and end up with a set of biomarkers. If I process the same data with my methodology, I'll isolate another set of biomarkers. And in every lab, it will most probably be another set.
I'm not saying it doesn't work, but there isn't enough harmonization and compatibility between what's done in one place versus another. The Metabolomics Standards Initiative and the Metabolomics Quality Assurance and Quality Control Consortium (mQACC) go in that direction, not only for GC×GC. But if we want to convince a clinician to use the technology, we have to be able to say that the data we produce and the outcomes of the research will be reproducible in any hospital – and that's not yet the case today.
That's where we are, and it still requires a lot of work.
Is this challenge unique to GC×GC?
I don't think it's unique to GC×GC, but I think the problem is even more significant for GC×GC because the quantity of data you're generating is huge. If you think about GC×GC coupled with high-resolution time-of-flight mass spectrometry (TOF-MS) in particular, the number of data points you're collecting, and the size of the data matrices you have to process, are tremendously bigger than with one-dimensional GC, for example, or even liquid chromatography.
The larger your data matrix, the more difficult it is to harmonize the way you're going to treat those data. The more data you have, the more options you have when applying statistics and everything else. That's where the challenge lies.
What are the main research priorities in your lab today?
We’re more focused now on volatile compounds. Medical applications probably account for 75 percent of what we do, but we'll tackle any analytical challenge involving volatile organic compounds (VOCs).
Sometimes we're contacted by industry – the food industry, the fragrance industry, or the petroleum industry. With the energy transition, for example, we have to characterize pyrolysis oils, which are very complex and very different from crude oils. We've also been working in archaeology, for example, to characterize very old pieces of hafting equipment. We still do some more theoretical modeling and model development to predict the use of chromatographic space. But breathomics, volatilomics, and lipidomics are the keywords in the lab right now.
The good thing is that everything we develop in one niche of application can be used in another. So an advance we make in data processing for the petroleum industry can be applied in the medical domain quite quickly afterward. That's really a good return on the investment in research.
What impact could GC×GC have in clinical medicine?
We're trying to improve our understanding of the chemical composition of patient samples. We know we can collect blood, urine, feces, or exhaled breath samples from patients. That's not new. But we're trying to get a more precise and exhaustive chemical description of what's in those samples to help identify a pathology – to distinguish between a cancer patient and a non-cancer patient, for example.
This is just one extra piece of the big puzzle that medical doctors have to solve. They already use genomics, proteomics, and a lot of clinical testing. Looking at VOCs and other volatile compounds provides an extra piece of information they can integrate into their assessment of the patient's overall condition. That's how I see our contribution.
There's also another opportunity at the screening level. The more patients you screen, the better your chances of identifying a pathology at an early stage.
We're working on ways to visually differentiate between the two-dimensional chromatogram of a patient and that of a control. How do those two images compare? Could we use that comparison for rapid screening and perhaps highlight some biomarkers?
But if you want to move this into hospitals, taking a GC×GC high-resolution TOF-MS system and putting it in a hospital isn't the way to go. It's still too complex. At this stage, the best approach is still to sample patients (exhaled air, micro-volumes of blood, and so on) inside the hospital infrastructure under strict procedures and then transfer samples to the analytical lab for analyses.
Instead, we can also use GC×GC in non-target screening mode to identify a few biomarkers and then transfer those findings to target techniques (e.g., 1D-GC-MS using specific sample preparation) if enough specificity can be ensured towards the targets.
Another option is to use proton-transfer-reaction mass spectrometry or selected-ion flow-tube mass spectrometry (SIFT-MS), in non-target mode, using multiple chemical precursors and specific ion channels. These are simpler, direct-inlet mass spectrometry techniques, which are much more suitable for hospitals. A SIFT-MS system is about the size of a dishwasher, and the patient can simply blow directly into it. That's what we're trying to do now. We've been duplicating studies using GC×GC-TOF-MS and SIFT-MS, and we're seeing very good correlations between the two.
Could GC×GC contribute to more personalized approaches to medicine?
For sure. The idea is to get a quick picture of an individual patient's condition. We're also doing more studies where we follow patients after treatment. For example, if a patient is diagnosed with cancer, we can monitor them over time. Are they responding to the treatment? Can we see differences in their volatilome that might help us better understand their response?
There's also growing interest in virtual biobanking where my colleague Pierre-Hugues Stefanuto is exploring how GC×GC datasets could contribute to creating high-resolution chemical descriptors of exhaled air samples rather than relying on the challenging long-termstorage of raw samples.
In general, when genomics, proteomics, and metabolomics are used in the framework of personalized medicine, we think that GC×GC-TOF-MS can contribute to that. It's something we'll be working on in the near future.
What other developments in GC×GC do you find particularly promising?
One important area of development is miniaturization, which could enable online measurements in chemical processes or the pharmaceutical industry.
We're seeing more valve modulators coming back now, with more robust designs that could allow us to make them smaller and smaller while maintaining good modulation efficiency for specific applications.
I think it's now possible to imagine miniaturized, robust GC×GC systems designed for particular applications. That's probably the next big thing – smaller systems that can be used continuously to monitor things like air quality for exposomics, for example, where we need more separation power than one-dimensional GC can provide.
Another important development will be the increasing integration of machine learning and artificial intelligence into data processing. This is typically a field where AI could help with data reduction, for example.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
