VariABEL (version 0.9-2.1)

var_test_gwaa: Function performs comparison of genotypic variances.

Description

The presence of a significant difference of genotypic variances points to possible presence of interaction between a tested SNP and a factor (or set of factors). This functions tests for that.

Usage

var_test_gwaa(formula, genodata, phenodata, genodata_info=NULL, testname="svlm", analysis_type="AAvsABvsBB")

Arguments

formula

Regression model used for analysis. In the first stage linear regression is run to exclude the main SNP effect. In this stage adjustment for covariates is performed.

genodata

The genotypes data in GenABEL format or as DatABEL object.

phenodata

The phenotypes data in format of data.frame object.

genodata_info

The file with SNP information (name, position). Used if genotype data is a DatABEL object.

testname

Name of the variance heterogeneity test to perform. "svlm" (for imputed genotype data), "Levene", and "Bartlett" test are supported. Note that "svlm" test only additive genetic model and the parameter analysis_type is ignored.

analysis_type

Type of analysis to perform. AAvsABvsBB - each henotypic group is testes against other two, AAvsABandBB - group AA tested against AB and BB, ABvsAAandBB - AB against AA and BB, BBvsAAandAB - BB against AA and AB. Only available for typed SNPs. Note that the input parameter analysis_type is ignored if testname="svlm"

Value

The output is a data.frame object. The table contains the chi^2 of the variance heterogeneity test (the name is chisq) the effects and standard errors of all covariates included into the regression model, main SNP effect (the names are snp_eff and snp_se). In the case of the svlm test the columns snp_eff_dispertion and snp_se_dispertion contain effect of a SNP on the squared values of the trait.

Details

The function var_test_gwaa tests for difference in genotypic variances. This difference points to the presence of possible interaction between the tested SNP and some factor. In the case of the svlm test the analysis consists of two stages: first the regular GWA is done where regression analysis is performed with covariates specified in the input parameter formula, in the second stage the regression analysis is performed with using residuals from the first stage and a SNP as a covariate.

References

Struchalin et al., Variance heterogeneity analysis for detection of potentially interacting genetic loci: method and its limitations. BMC Genetics 2010, 11:92, doi:10.1186/1471-2156-11-92

Struchalin et al., An R package VariABEL for genome-wide searching of potentially interacting loci by testing genotypic variance heterogeneity. Submitted.

Examples

if (require(GenABEL)) {
    data(srdta)
    result1 <- var_test_gwaa(bt ~ qt1 + qt2,
                             genodata=gtdata(srdta),
                             phenodata=phdata(srdta))

    # If there is covariates are needed:
    result2 <- var_test_gwaa("bt",
                             genodata=gtdata(srdta),
                             phenodata=phdata(srdta))
}