Statistical contributrion to the analysis of metabonomic data in 1H-NMR spectroscopy

Rousseau, Réjane
(2011)

Files

TheseRéjaneRousseau.pdf
  • Open Access
  • Adobe PDF
  • 4.2 MB
Download

Details

Authors
  • Rousseau, RéjaneUCLouvain
    author
Supervisors
Govaerts, Bernadette
 ;
Verleysen, Michel
Abstract
(en) This thesis provides a complete methodology to lead a metabonomic study from the data acquisition to the discovery of biomarkers. Metabonomics aims to discover biomarkers or the metabolites altered in biological reactions. 1H-NMR spectroscopy is used to generate spectral profiles describing the composition of metabolites in collected biofluid samples. Statistical comparisons of spectra obtained in various biological states permits a quantitative investigation of metabolites changes inherent to a studied biological reaction. The enormous amount of metabonomic data lets foreseeing opportunities for biological and pathological mechanisms discoveries as for clinical diagnosis and for solutions to speed up the drug discovery process. The use of statistics in metabonomics is commonly restricted to the search for biomarkers in the data. However, the success of a metabonomic study does not only depend on the choice of a statistical data analysis method but also on the information contained in the data. From this perspective, the methodology presented in this thesis brings statistical contributions to metabonomic on two axes. On one hand statistical tools are developed to improve the quality of the data. Databases and methods evaluating the variability of spectra are used to define an adequate mode of sample collection, preparation, 1H-NMR measurement. An automated 1H-NMR data pre-treatment procedure offers solutions to decrease instrumental and biological information out of interest. On the other hand, statistical methods to search for biomarkers in metabonomic data are developed and evaluated: a battery of six methods for two-class studies and a four steps methodology combining ICA and statistical modelling for two- and multiple-class as quantitative studies. The whole methodology is successfully applied to a real metabonomic study for the search of AMD disease biomarkers.
Affiliations

Citations

  • APA
  • Chicago
  • FWB

Rousseau, R. (2011). Statistical contributrion to the analysis of metabonomic data in 1H-NMR spectroscopy. https://hdl.handle.net/2078.5/208973