A Statistical Association Test for the Identification of Clustered Disease Risk Variants

Abstract : Genome-wide association studies have identified numerous common variants associated with a wide variety of complex diseases. However these variations only explain a small proportion of the heritability. A hypothesis is that rare variants may play an important role in disease risk. Testing association between rare variants and diseases represents a challenge due to their very low frequency in the population. Statistical methods have been developed testing the association with group of rare variants, since CAST (Cohort Allelic Sum Test) described in 2007 by Morgenthaler and Thilly. Few of these tests (Fier et al., 2012; Ionita-Laza et al., 2012; Lin, 2014; Schaid et al., 2013) take into account spatial genetic information. However, it is has been shown that disease variants may cluster in important functional domains of genes. For instance, pathogenic mutations are localized in the gene FLNA, causing congenital malformations (Robertson et al., 2003). Ionita-Laza et al. also identified in 2012, clusters of rare disease variants in the gene LRP2, associated with autism spectrum disorders. We developed a statistical test, DoEstRare, whose aim is detecting clusters of disease rare variants while preserving sufficient power when there is no cluster but still enrichment of rare alleles (overall the sequence). The DoEstRare statistics consists in comparing the mutation distributions, estimated by kernel method, between cases and controls. We compared the power and the type I error of our method to several published association tests for rare variants. Power and type I error computations are based on simulations conducted under two main genetic scenarios: absence or presence of one cluster of causal variants, varying also the proportion of causal variants. We are also now simulating with the software COSI according to the model developed in Schaffner et al. (2005) in order to mimic the demographical history of the European population. We observed the change in power according to the introduction of different statistical components. This test is thus adapted to non-cluster situations with the use of a burden component. A weighting scheme is also adopted in order to better discriminate between causal and neutral variants. We show consistent increase in power in both scenarios (for X % up to39.7% increase at best compared to SKAT-O). While this may be specific of the simulations carried out here, we think that DoEstRare represents a convenient and powerful alternative to test rare allele variants effects when there is no prior hypothesis of the real distribution of causative alleles.
Type de document :
Communication dans un congrès
. European Mathematical Genetics Meeting (EMGM), Apr 2015, Brest, France. 79, pp.43-44, 2015, Human Heredity
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01222285
Contributeur : Lise Bellanger <>
Soumis le : jeudi 29 octobre 2015 - 15:58:31
Dernière modification le : lundi 25 juin 2018 - 09:33:47

Identifiants

  • HAL Id : hal-01222285, version 1

Collections

Citation

Elodie Persyn, Matilde Karakachoff, Floriane Simonet, Jean-Jacques Schott, Richard Redon, et al.. A Statistical Association Test for the Identification of Clustered Disease Risk Variants. . European Mathematical Genetics Meeting (EMGM), Apr 2015, Brest, France. 79, pp.43-44, 2015, Human Heredity. 〈hal-01222285〉

Partager

Métriques

Consultations de la notice

155