Skip to Main content Skip to Navigation

Évaluation de modèles pronostiques issus de l'analyse du

Abstract : The challenge faced in studying gene expression profiles with cDNA microarrays is to propose new
prognostic biomarkers. However, the need for a simultaneous analysis of thousands of genes in a
limited number of patients leads to several statistical problems. We studied two aspects related
to the evaluation of prognostic models stemming from transcriptome analysis. In a first step, the
complementary use of simulated and real public datasets allowed us to show that the choice of
the most appropriate method depends on the extent to which the theoretical properties of that
method adapt to the data structure. This study was carried out using discriminant and betweengroup
analyses. In a second step, simulations allowed us to estimate the respective contributions
of classical clinical variables and transcriptomic variables in using survival models. In addition, we
identified the parameters associated with overestimation of microarray contribution.
Document type :
Complete list of metadatas

Cited literature [156 references]  Display  Hide  Download
Contributor : Caroline Truntzer <>
Submitted on : Tuesday, July 10, 2007 - 9:09:59 AM
Last modification on : Wednesday, November 20, 2019 - 2:57:50 AM
Long-term archiving on: : Thursday, April 8, 2010 - 8:00:22 PM


  • HAL Id : tel-00161161, version 1



Caroline Truntzer. Évaluation de modèles pronostiques issus de l'analyse du
transcriptome. Sciences du Vivant [q-bio]. Université Claude Bernard - Lyon I, 2007. Français. ⟨tel-00161161⟩



Record views


Files downloads