Abstract
Volatile metabolites in exhaled air have promising potential as diagnostic biomarkers. However, the combination of low mass, similar chemical composition, and low concentrations introduces the challenge of sorting the data to identify markers of value. In this paper, we report the development of pyAIR, a software tool for searching for volatile organic compounds (VOCs) markers in multi-group datasets, tailored for Thermal-Desorption Gas-Chromatography High Resolution Mass-Spectrometry (TD-GC-HRMS) output. pyAIR aligns the compounds between samples by spectral similarity coupled with retention times (RT), and statistically compares the groups for compounds that differ by intensity. This workflow was successfully tested and evaluated on gaseous samples spiked with 27 model VOCs at six concentrations, divided into three groups, down to 0.3 nL/L. All analytes were correctly detected and aligned. More than 80% were found to be significant markers with a p-value < 0.05; several were classified as possibly significant markers (p-value < 0.1), while a few were removed due to background level. In all group comparisons, low rates of false markers were found. These results showed the potential of pyAIR in the field of trace-level breathomics, with the capability to differentially examine several groups, such as stages of illness.