Development of a statistical framework for mass spectrometry data analysis in untargeted Metabolomics studies
by Alexander Kaever
Date of Examination:2014-06-06
Date of issue:2014-12-11
Advisor:Prof. Dr. Burkhard Morgenstern
Referee:Prof. Dr. Burkhard, Morgenstern
Referee:Prof. Dr. Ivo Feussner
Files in this item
Name:dissertation.pdf
Size:4.85Mb
Format:PDF
Abstract
English
A central objective in the analysis of mass spectrometry-based untargeted Metabolomics data is the detection of intensity patterns that differ between experimental conditions and the identification of underlying metabolites and biochemical processes. In this context, the identification of metabolites is a major bottleneck and needs to be guided by expert knowledge and tools for explorative data analysis. The integration of data sets from other omics platforms, e.g. DNA microarray-based Transcriptomics, can thereby provide valuable hints and support the reconstruction of related metabolic pathways, which then form the biochemical context for metabolite identification. In this work, a statistical framework and user interfaces for exploratory evaluation of mass spectrometry-based non-targeted Metabolomics data in combination with data sets from other omics platforms are introduced. The developed methods and tools were combined in the highly interactive MarVis-Suite software. The MarVis-Filter interface includes functions for the adduct and isotope correction of mass spectrometry data, molecular formula prediction, statistical ranking, filtering, and combination of cross-omics data sets. Within MarVis-Cluster, intensity profiles associated with ion species or microarray spots (features) in filtered and combined data sets can be clustered, visualized, interactively inspected and labeled. By means of MarVis-Pathway, data set features may be annotated in the context of organism-specific metabolic pathways. For statistical analysis, which forms a counterweight to the highly interactive and selective MarVis workflow, an extensive framework for meta-analysis of multi-omics data sets based on pathway enrichment analysis was developed. The methods and tools were successfully applied to several liquid chromatography/mass spectrometry data sets in combination with DNA microarray data in the context of plant wounding. The integration of Transcriptomics data thereby significantly supported the analysis and interpretation of non-targeted Metabolomics data sets.
Keywords: Bioinformatics; Metabolomics; Metabolic fingerprinting; Mass spectrometry; Metabolic pathways; Set enrichment analysis; Transcriptomics; Meta-analysis