If you’ve used gas chromatography-mass spectrometry (GC-MS) to analyze unknown compounds from cells and biological tissue, a new study suggests you may want to throw away most—if not all—of your data. But some researchers believe it might be better to keep the data and throw away the new study instead.

Gary Siuzdak of Scripps Research Institute California and coworkers heated standard samples of known small molecules and samples of unknown metabolites from cells to mimic the GC stage of a GC-MS instrument. This initial stage uses heating to volatilize and separate samples. The researchers then used liquid chromatography-MS (LC-MS), which does not use heat in its initial LC separation stage, to analyze the heated samples as well as unheated, but identical, samples (Anal. Chem. 2015, DOI: 10.1021/acs.analchem.5b03003). The result: Up to 40% of the heated compounds were modified or destroyed, compared with the unheated ones—even in samples whose components were derivatized by trimethylsilylation, a method widely used to protect compounds from heating.

The results question what we are really analyzing when GC-MS experiments are used to identify unknown compounds in a sample, says Siuzdak, a specialist in MS and metabolomics. “The injected sample, or its thermal degradation products?” he asks.

“We found that even relatively low temperatures used in GC-MS can have a detrimental effect on small-molecule analysis,” he continues. GC and GC-MS have been used to identify and measure small molecules for more than 50 years. So the new results might trigger an “uh-oh moment” for analytical chemists.

Using standard laboratory procedures, Siuzdak and coworkers used LC-MS to analyze small-molecule standards and human plasma metabolites stored at room temperature (25 °C). Then they compared the results with LC-MS analyses of the same samples heated to 60 °C, 100 °C, or 250 °C, to mimic a variety of common heating conditions used in GC-MS. The heating changed the identities of or destroyed roughly 5%, 15%, and 30% of compounds in both types of samples, respectively.

Siuzdak suggests that work-arounds would involve using GC-MS primarily to measure sample concentrations by using reference standards rather than trying to identify unknowns, or switching to techniques such as LC-electrospray ionization MS, which does not use heating. “In fact, that is the direction things are going,” he says. “However, tens of thousands of GC-MS instruments are still being used on a daily basis” to analyze unknowns in nutrition, forensics, clinical and environmental analysis, and similar fields.

The findings have sparked some strong opinions, both heated and unheated. For example, “I am unclear why a scientific journal would publish work that is such clear nonsense,” says metabolomics expert Oliver Fiehn of the University of California, Davis. “The publication of this paper is, in my opinion, a major embarrassment for Analytical Chemistry. There was a peer review process involved, but perhaps not all peer reviewers understood the study design the Siuzdak group used.”

When analytical chemists “develop, validate, and implement an analytical method for GC-MS, they carefully control for important method parameters,” such as the temperatures used and the conditions used to derivatize compounds to protect them from heating, he says, noting that the Siuzdak group sidestepped such controls.

In some experiments, Siuzdak’s team “used underivatized metabolites and heated them intensely,” Fiehn says. “That is called cooking—like in a kitchen. Primary metabolites have lots and lots of hydroxyl and amino groups, and blood plasma has a lot of sugars,” groups that should be protected to prevent breakdown before heating them for analysis, he adds.

Fiehn believes the Scripps researchers made other methodological mistakes, such as analyzing trimethylsilyl-derivatized samples in a water-containing LC-MS solvent, because water cleaves trimethylsilyl groups. “You cannot inject trimethylsilylated compounds in an aqueous solvent into an LC-MS system and expect that peaks survive,” he says. “That’s why their MS spectra could not identify compounds: The authors destroyed them.”

And metabolomics specialist Warwick Dunn of the University of Birmingham, in England, pointed out that “the published study heated dry samples, whereas GC heats samples in a liquid solution for a few seconds during injection and then in the gas phase. Heat transfer in a solid can be expected to be higher and could therefore provide a higher level of degradation than in the gas phase.” Hence, “further validation of the degradation of many metabolites is required before we should worry about the terabytes of data already collected.”

Other scientists who spoke with C&EN were less critical. For example, Stephen Barnes, director of the Targeted Metabolomics & Proteomics Laboratory at the University of Alabama, Birmingham, says that whole-metabolome analysis typically identifies fewer than 20% of cellular metabolites, and the new study could help explain why.

At this year’s International Conference of the Metabolomics Society, Barnes says, “a group reported incubating a series of pure metabolites at 70 °C and analyzing them to see if new compounds appeared without added enzymes. The answer was yes.” He notes that compounds in samples can react with each other, albeit slowly, and that any elevated temperatures to which they are exposed can also alter their composition.

“I don’t think the problem is confined to GC-MS,” Barnes adds. “Those doing LC-MS analyses need to think hard about this too.” For example, samples can get modified during the extraction process used to prepare them, in which heated solvents are sometimes used. Or they can even get modified at the source from which they are obtained—for example, the human body is about 37 °C. GC-MS and LC-MS analyses are used as evidence in criminal cases, Barnes says, and “it seems like one could defend guilty people on the basis that prosecutors have no way of knowing whether the data are valid. The forensic science community needs to develop a more rigorous understanding of the pitfalls of analysis.”

Liang Li of the University of Alberta, whose group develops MS methods for proteomics and metabolomics applications, says the new study is important because it reminds people, particularly new practitioners in the metabolomics field, not to use high temperatures or other conditions that are more extreme than necessary for processing complex metabolomic samples.

Siuzdak says his team’s study wasn’t intended to disparage “thousands of papers’ worth of research.” But molecular transformation from sample heating “has been a fundamental yet unrecognized problem with GC-MS technology since its inception,” he says. “I remember asking someone about it many years ago during the question period after his talk, and he simply didn’t know how to respond. For me, it has always been the elephant in the room.”