Combination of Independent Component Analysis and statistical modelling for the identification of metabonomic biomarkers in 1H-NMR spectroscopy
(2008) , 28 pages
Files
ISBADP0941.pdf
Open Access - Adobe PDF
- 10.01 MB
Details
- Authors
Govaerts, Bernadette 
- Abstract
- In order to maintain life, living organisms product and transform small molecules called Metabolites. Metabonomics is a scientific platform studying the development of Biological reactions caused by a contact with a physio-pathological stimulus through the metabolites. me The 1H-NMR spectroscopy is used to describe the metabolites composition on the basis of spectra. Biologists can then confirm the development of a biological reaction if specific spectral regions (« biomarkers ») are altered in spectra obtained in given physiological situations. However, this process supposes a preliminary identification in an experimental database of the biomarkers or spectral regions to examine because of their changes in case of the biological response. Traditionally, this identification is realised with some limitations, by examination of the 2 first components from a Principal Component Analysis. This paper presents a new methodology in four steps providing two kind of knowledge on 1H-NMR metabonomics biomarkers: the identification of biomarkers and the visualization of the effects on the biomarkers caused by external changes of interest. A first step employs Independent Component Analysis in order to decompose the spectral data into statistically independent components or sources. The independent pure or composite metabolites contained in the studied biofluid are discovered through the sources and their quantity through the mixing weights. The advantages of independent components to overview the data are described comparatively to the usual PCA analysis. Solution for questions specific to ICA like the choice of the number of components and their ordering have been developed. The second step consist on a statistical modelling applied to the ICA results. Statistical hypothesis tests on the parameters of the estimated models lead in the third step, to select sources presenting biomarkers or spectral regions changing significantly according to the factor of interest. A panel of various statistical models is considered adaptively to the possible the nature of the biomarker question. Finally, the last step proposes the computation of contrasts to visualize changes on the spectral biomarkers caused by different changes of factor interest. The methodology and its efficiency is illustrated on two experimental datasets.
- Affiliations
Citations
Rousseau, R., Govaerts, B., & Verleysen, M. (2008). Combination of Independent Component Analysis and statistical modelling for the identification of metabonomic biomarkers in 1H-NMR spectroscopy (STAT Discussion Paper 0941). https://hdl.handle.net/2078.5/24382