Principal Investigator: Dr Edith Le Floch
CEA, Evry, FranceTags: 45408, complex phenotypes, copy-number variants, GWAS, heritability, Machine Learning
Many associations have been identified by Genome-Wide Association Studies (GWAS) between Single Nucleotide Polymorphisms (SNPs) and complex phenotypes. However, these variants only explain a limited amount of the genetic variability of these phenotypes.
Several hypotheses have been suggested to explain this missing heritability such as the additivity of many weak SNP effects not detected in GWAS, which has been partially confirmed by several works on the heritability of complex phenotypes, or the implication of rare variants. Another explanation could be that an important part of the relevant genetic information lies in Copy Number Variations (CNVs) rather than in SNPs.
A large number of papers show that there exists a link between CNVs and some neurodevelopmental diseases like schizophrenia or autism. However, few papers report associations between CNVs and other diseases.
We thus propose to investigate the CNVs involved in complex phenotypes in a similar way to SNPs in GWAS and heritability estimation.
Moreover, we also propose a joint analysis of SNPs and CNVs for heritability estimation and phenotype prediction using machine learning approaches, in order to both assess the proportion of phenotypic variance that can be explained by genetic variants and identify these variants.
We will return to the community the developed methods that allow analyzing simultaneously CNVs and SNPs in the form of articles and programming code (R packages). In addition, we will return a list of significant CNV regions, SNPs, and genes associated with complex diseases or traits, as well as the estimations of global heritabilities.
We will expect to observe a better prediction of the disease status/trait than with classical genetic analyses that use only SNP data.