Uncovering Hidden Compositional Changes in Breath Profiles using Thermal Desorption with GC×GC-TOF MS and Chemometrics

Terry Buckland, Markes International, Bridgend, UK (tbuckland@schauenburganalytics.com)
Laura McGregor, Sepsolve Analytical, Peterborough, Uk
Anthony Buchanan, Sepsolve Analytical, Peterborough, Uk
Bob Green, Sepsolve Analytical, Peterborough, Uk
Caroline Widdowson, Markes International , Bridgend, Uk

Volatile organic compounds (VOCs) emitted in breath have great potential for use in non-invasive disease diagnosis. This is largely due to the discovery of so-called ‘biomarkers’, which provide indicators of normal or abnormal states.

In large-scale clinical trials, hundreds of samples may be collected across multiple sites over the course of many weeks. During this biomarker discovery phase, an incorrect identification can compromise the validity of an entire trial, meaning that both robust analytical techniques and confident data mining are required.

Thermal desorption (TD) coupled with GC–MS is known as the ‘gold standard’ for breath analysis, due to its ability to capture a complete breath profile with high sensitivity. However, breath can contain hundreds of different VOCs, often in trace levels - making it difficult to isolate and identify biomarkers of disease.

Here, we will show how combining TD with comprehensive two-dimensional gas chromatography and time-of-flight mass spectrometry (GC×GC-TOF MS) can provide enhanced separation of these complex samples to uncover hidden biomarkers.

However, data acquisition is just the beginning – the information-rich chromatograms must then be transformed into meaningful results. Here, we demonstrate the use of a powerful data mining and chemometrics platform to automatically find the significant differences in complex datasets and to create statistical models to predict the class of future samples.

Firstly, chromatographic alignment accounts for retention time drift over the course of the study and minimises the risk of false hits. Feature discovery is then performed on the raw data to find significant changes across sample classes. In metabolomics matrices, the diagnostic compounds are rarely of high abundance - by adopting a raw data approach, trace differences are less likely to be overlooked.