Development of computational methods for metabolic network analysis based on metabolomics data

Abstract

The baker';s yeast, Saccharomyces cerevisiae, is a simple eukaryotic organism with approximately 6000 genes. Saccharomyces cerevisiae is an ideal model organism for large-scale functional studies and provides a system in which genes can be systematically inactivated by way of gene-knockout methods. A substantial fraction of the 6000 genes in Saccharomyces cerevisiae encode proteins for which currently we do not know any confirmed or putative function. Prediction of the functional role of these proteins is a challenging problem in systems biology, especially as many of these genes have no overt phenotypes. In our study, we aim at a better understanding of the underlying functional relationships between genes working across diverse metabolic pathways using intracellular metabolite profiling studies. We applied bioinformatics methods and statistical analysis techniques in combination with metabolic profiling to understand the function and the regulatory mechanisms of specific genes involved in central carbon metabolism and amino acid biosynthesis. The experimental work was carried out by the group of Prof. Elmar Heinzle (Biochemical Engineering, Saarland University), our collaboration partner. 13C stable isotope substrates can be used as tracers to generate detailed metabolic profiles of gene knockouts. Detailed and quantitative information on the physiological cellular states is measured by 13C -metabolic profiling of cultures grown on novel high throughput oxygen sensor microtiter plates. In this dissertation, we worked towards developing systematic approaches for study of Saccharomyces cerevisiae genes of unknown function based on the metabolic profiles of knockout mutants under varied environmental conditions. In the first step, we have developed a software tool called CalSpec for automation of Gas Chromatography Mass Spectrometry data acquisition and analysis routine, as this is a bottleneck in the metabolic profiling studies. In the next step, we worked on large scale statistical analysis of metabolic profiling data. We applied various algorithms for finding closely related mutants which show similar metabolic profiles. According to our hypothesis, similarity in the metabolic profiles can be used to find functionally linked genes. Saccharomyces cerevisiae is known to be robust to majority of genetic perturbations. In these cases where the mutants show no overt 4 phenotypes, we developed a sensitive outlier detection method to detect those subsets of metabolic profile features which are most differentiating (outliers) for all mutants. The second part of this dissertation involves developing computational tools for metabolic pathway analysis on the basis of genome scale metabolic models, as well as integration of various newly emerging experimental techniques. In recent years, genome scale metabolic models have been and are continuing to be assembled for various organisms. In the year 2003, first comprehensive genome scale metabolic model for yeast became publicly available. With the emergence of system biology area of research, diverse computational approaches have been developed. In this work, we developed a new webserver called MetaModel, for analysis of genome scale metabolic networks of eukaryotic organisms.