Discovery based yeast metabolomic analysis using comprehensive two-dimensional gas chromatography with time-of-flight mass spectrometry and chemometrics

Abstract

Comprehensive two-dimensional (2D) separations are highly beneficial in the characterization of complex samples. The research presented in this dissertation involves the analysis of the small polar molecules comprising the yeast metabolome using commercially available GCxGC-TOFMS instrumentation. The studies described herein are some of the first on metabolomic data using GCxGC-TOFMS and the analysis required a tremendous amount of procedural and software method development. A proof of principle experiment was performed using fermenting and respiring yeast cells. Initially, principal component analysis (PCA) was performed on three selective mass channels (m/z ) to identify the metabolite locations exhibiting changes between sample types. Twenty-six metabolite peaks were reported. Following proof that the GCxGCTOFMS is applicable to the study of yeast metabolite data, an extensive study was performed to determine the ability of GCxGC-TOFMS to detect metabolites and ultimately be utilized to distinguish classes. In this study, a newly developed in-house Fisher ratio method was applied, using the entire 3D data cube (i.e., column 1 retention time, column 2 retention time, m/z), more than doubling the number of quantified metabolite peaks, relative to the PCA study. To obtain the confident identification and accurate metabolite quantification, parallel factor analysis (PARAFAC) was applied. The statistical significance of the results was determined by applying a Student's t-test to the deconvoluted peak volumes. It was determined that 54 identified metabolite peaks were statistically different at the 95% confidence level.Some cell strains exhibit robust oscillations in gene expression and molecular oxygen. A study with metabolites collected from cells showing periodicity in molecular dissolved oxygen was also performed to ascertain the changes in the metabolome. A new software method was developed to calculate the severity of the periodic pattern without signal bias or limiting the cycling frequency. This method is based on the raw signal intensities and, although it searches all m/z, only information from the three most selective m/z is output. Peak volumes were determined by PARAFAC and objective grouping of the metabolites with similar patterns was obtained with PCA. Metabolites were shown to cycle with four different patterns, with nineteen cycling with similar frequency but ∼180° out of phase.