Skip to Main content Skip to Navigation

Analyse statistique des données issues des biopuces à ADN

Abstract : This dissertation is dedicated to the statistical analysis of microarray data. We consider three issues linked to the transcriptome data.

In the first chapter, we study the problem of data normalisation; its purpose is to eliminate the parasite differences between samples, so as to retain only those variations that are due to biological phenomena. We present several existing normalisation methods and we propose improvements for some of them. Furthermore, in order to guide the choice among those methods, we develop a procedure to simulate microarray data.

In the second chapter, we deal with the detection of differentially expressed genes between two series of experiments, an issue that we assimilate to a multiple hypothesis testing problem. Several approaches are studied\string: model selection and penalty, FDR method based on a wavelet decomposition of the test statistics and Bayesian thresholding.

In the last chapter, we consider the problem of supervised classification of microarray data. To cope with the high-dimensionality issue, we develop a semiparametric method for dimension reduction, based on the maximisation of a local likelihood criterion in generalized linear single-index models. The dimension reduction step is then followed by a local polynomial regression step, in order to perform the supervised classification of the given individuals.
Document type :
Complete list of metadata
Contributor : Julie Peyre <>
Submitted on : Sunday, March 26, 2006 - 3:24:24 PM
Last modification on : Wednesday, March 10, 2021 - 1:50:03 PM
Long-term archiving on: : Saturday, April 3, 2010 - 11:02:48 PM


  • HAL Id : tel-00012041, version 1




Julie Peyre. Analyse statistique des données issues des biopuces à ADN. Mathématiques [math]. Université Joseph-Fourier - Grenoble I, 2005. Français. ⟨tel-00012041⟩



Record views


Files downloads