A multiple-method comparative study using GC-MS, AMDIS and in-house-built software for the detection and identification of "unknown" volatile organic compounds in breath.

The human respiratory system is a highly complex matrix that exhales many volatile organic compounds (VOCs). Breath-exhaled VOCs are often "unknowns" and possess low concentrations, which make their analysis, peak digging and data processing challenging. We report a new methodology, applied in a proof-of-concept experiment, for the detection of VOCs in breath. For this purpose, we developed and compared four complementary analysis methods based on solid-phase microextraction and thermal desorption (TD) tubes with two GC-mass spectrometer (MS) methods. Using eight model compounds, we obtained an LOD range of 0.02-20 ng/ml. We found that in breath analysis, sampling the exhausted air from Tedlar bags is better when TD tubes are used, not only because of the preconcentration but also due to the stability of analytes in the TD tubes. Data processing (peak picking) was based on two data retrieval approaches with an in-house script written for comparison and differentiation between two populations: sick and healthy. We found it best to use "raw" AMDIS deconvolution data (.ELU) rather than its NIST (.FIN) identification data for comparison between samples. A successful demonstration of this method was conducted in a pilot study (n = 21) that took place in a closed hospital ward (Covid-19 ward) with the discovery of four potential markers. These preliminary findings, at the molecular level, demonstrate the capabilities of our method and can be applied in larger and more comprehensive experiments in the omics world.