Skip to Main content Skip to Navigation
Theses

Statistical modelling for differential gene expression studies: variance-covariance models, sequential and meta-analysis.

Abstract : Microarrays enable to simultaneously study gene expression levels from several thousands of genes with very few samples. Three approaches are considered in this PhD work in order to overcome sensitivity problems in differential gene expression studies: variance-covariance modelling, sequential and metaanalysis. The first and the third parts mainly rely on shrinkage approaches, which consist in estimating each individual gene value by taking into account information from all genes of the experiment. By decreasing the total number of parameters to estimate, this increases sensitivity, that is to say the proportion of true positives among the truly differentially expressed genes. While variance modelling is always useful with small sample size designs, covariance modelling is especially important in time course studies where measures are repeated on the same individuals. Concerning sequential analysis, sensitivity is studied as a stopping rule. The aim is to stop the experiment before the scheduled end as soon as this criterion is higher than a given threshold, which enables to decrease costs. Meta-analysis is then studied in a wider context than sequential analysis where intermediate analyses were combined. It increases sensitivity by gathering results from individual studies, for which a direct comparison would be impossible, but answering the same biological question. Meta-analysis is studied both from the frequentist (effect size and p-value combinations) and the bayesian points of view.
Document type :
Theses
Complete list of metadatas

https://pastel.archives-ouvertes.fr/tel-00458988
Contributor : Guillemette Marot <>
Submitted on : Monday, February 22, 2010 - 10:21:34 PM
Last modification on : Monday, October 19, 2020 - 11:09:30 AM
Long-term archiving on: : Friday, June 18, 2010 - 9:40:29 PM

Identifiers

  • HAL Id : tel-00458988, version 1
  • PRODINRA : 248387

Collections

Citation

Guillemette Marot. Statistical modelling for differential gene expression studies: variance-covariance models, sequential and meta-analysis.. Life Sciences [q-bio]. AgroParisTech, 2009. English. ⟨NNT : 2009AGPT0039⟩. ⟨tel-00458988⟩

Share

Metrics

Record views

3989

Files downloads

1728