Skip to Main content Skip to Navigation
Theses

Detection of epistasis in genome wide association studies with machine learning methods for therapeutic target identification

Abstract : By offering an unprecedented picture of the human genome, genome-wide association studies (GWAS) have been expected to fully explain the genetic background of complex diseases. So far, the results have been mitigated to say the least. This, among other things, can be partially attributed to the adopted statistical methodology, which does not often take into account interaction between genetic variants, or epistasis. The detection of epistasis through statistical models presents several challenges for which we develop in this thesis a pair of adequate tools. The first tool, epiGWAS, uses causal inference to detect epistatic interactions between a target SNP and the rest of the genome. The second tool, kernelPSI, instead uses kernel methods to model epistasis between nearby single-nucleotide polymorphisms (SNPs). It also leverages post-selection inference to jointly perform SNP-level selection and gene-level significance testing. The developed tools are -- to the best of our knowledge -- the first to extend powerful statistical learning frameworks such as causal inference and nonlinear post-selection inference to GWAS. In addition to the methodological contributions, a special emphasis was placed on biological interpretation to validate our findings in multiple sclerosis and body-mass index variations.
Document type :
Theses
Complete list of metadatas

Cited literature [252 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/tel-02895919
Contributor : Abes Star :  Contact
Submitted on : Friday, July 10, 2020 - 11:16:34 AM
Last modification on : Wednesday, October 14, 2020 - 4:11:15 AM

File

2020UPSLM006_archivage.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02895919, version 1

Citation

Lotfi Slim. Detection of epistasis in genome wide association studies with machine learning methods for therapeutic target identification. Quantitative Methods [q-bio.QM]. Université Paris sciences et lettres, 2020. English. ⟨NNT : 2020UPSLM006⟩. ⟨tel-02895919⟩

Share

Metrics

Record views

111

Files downloads

70