Metabolomics is the study of low molecular weight molecules found within cells and biological systems. It has emerged for deciphering the complex time-related concentration, activity and flux of metabolites in biological samples. Multiple analytical platforms such as liquid chromatography-mass spectrometry, gas chromatography-mass spectrometry, and nuclear magnetic resonance spectroscopy have been used in metabolomics. Unlike other analytical platforms, the comprehensive two-dimensional gas chromatography mass spectrometry system provides a much increased separation capacity, chemical selectivity and sensitivity for the analysis of metabolites. The aim of this project is to infer metabolic networks from the measurements of metabolites using this high-throughput comprehensive two-dimensional gas chromatography mass spectrometry system. To achieve a high accuracy metabolic network construction, an innovative peak detection algorithm will be first developed for simultaneous baseline removal, smoothing and peak picking steps, as well as the statistical assessment of peak confidence level. The investigators will then develop an innovative peak alignment algorithm based on global comparison using point matching algorithm as well as considering measurement error and outlier detection. Once an accurate alignment peak table is created, metabolic networks will be created using Bayesian ensemble method, which is Bayesian model averaging, to find meaningful pairwise interactions and associations among sets of metabolite peaks as well as to account for uncertainty of network complexity. In particular, the pathway information will be incorporated into network constructions for weighting each network constructed by several different approaches.
The comprehensive two-dimensional gas chromatography mass spectrometry platform can not only leverage the peak capacity issue in current metabolomics, but also brings much improved chemical selectivity and sensitivity for metabolomics analyses. The developed statistical and computational approaches will lead to new methodology for data pre-processing and metabolic network construction for the analysis of high-throughput data. The metabolic network construction system developed from this project will enable identification of consistent metabolite mechanisms and pathways for numerous organisms and diseases in real time. Thus this research will have broader impacts for biomedical research, medicine and public health. The methods developed in this research can also be applied to the analysis of liquid chromatography and mass spectrometry based metabolomics data.
