Human axillary sweat is a poorly explored biofluid within the context of metabolomics when compared to other fluids such as blood and urine. In this paper, we explore the volatile organic compounds emitted from two different types offabricsamples (cotton and polyester) whichhad beenworn repeatedly during exercise by participants. Headspace solidphase microextraction (SPME) and comprehensive twodimensional gas chromatography time-of-flight mass spectrometry (GC×GC-TOFMS) were employed to profile the (semi)volatile compounds on the fabric. Principal component analysis models were applied to the data to aid in visualizing differences between types of fabrics, wash treatment, and the genderofthesubjectwhohadwornthefabric.Statisticaltools included with commercial chromatography software (ChromaTOF) and a simple Fisher ratio threshold-based feature selection for model optimization are compared with a custom-written algorithm that uses cluster resolution as an objective function to maximize in a hybrid backwardelimination forward-selection approach for optimizing the chemometric models in an effort to identify some compounds that correlate to differences between fabric types. The custom algorithm is shown to generate better models than the simple Fisher ratio approach.
A. Paulina de la Mata, PhD MSc MBA, UDLAP.
Rachel H. McQueen
Seo Lin Nam
James J. Harynuk