alpha1 is set to a value smaller than 1, then the two-step procedure of
Gauderman et al. (2010) will be used to first select all SNPs showing a p-value smaller than alpha1
in a logistic regression of the environmental factor against the sums of the codings for the parents' genotypes at the respective
SNP. In the second step, the genotypic TDT is then applied to the selected SNPs.If unstructured = TRUE, all fully parameterized model is considered and a likelihood ratio test is performed.
While colGxE computes the p-values based on asymptotic ChiSquare-distributions,
colGxEPerms can be used to determine permutation-based p-values for the basic genotypic TDT (i.e. for colGxE
using alpha = 1 and unstructured = FALSE.
colGxE(mat.snp, env, model = c("additive", "dominant", "recessive"), alpha1 = 1, size = 50, addGandE = TRUE, whichLRT = c("both", "2df", "1df", "none"), add2df = TRUE, addCov = FALSE, famid = NULL, unstructured = FALSE)
colGxEPerms(mat.snp, env, model = c("additive", "dominant", "recessive"), B = 10000, size = 20, addPerms = TRUE, famid = NULL, rand = NA)NA.
This matrix might be generated from a ped-file by, e.g., employing ped2geno.
mat.snp) containing for each offspring the value of a binary environmental
variable, which must take the values 0 and 1.
model = "dom" will
fit a dominant model, and model = "r" an recessive model.
alpha1 = 1, all SNPs will be tested with a genotypic TDT.
Otherwise, the two-step procedure of Gauderman et al. (2010) will be used to select all SNPs showing a p-value smaller than
or equal to alpha1 in a logistic regression in which the environmental factor is used as response and the
sums over the codings for the genotypes of the parents are employed as predictor. The genotypic TDT will then be applied to
the selected SNPs. Since a logistic regression is employed in the first step, which requires a numerical determination of
the parameter estimates, the two-step procedure will not lead to a reduction in computing time, but will increase the computing time.
"2df", 2 degree of
freedom likelihood ratio tests comparing the fitted models (containing one parameter for the SNP and one for the gene-environment
interaction) with models containing no factor will be performed. If "1df", one degree of freedom likelihood ratio
tests comparing the fitted model (containing two parameters, one for the SNP and the other for the interaction) with models
only containing the respective SNP will be added to the output. If "both" (default), both tests will be performed,
whereas none test will be done, if whichLRT = "none".addCov = FALSE, as this covariance is given by the negative variance of the parameter estimate
for the SNP.env specifying the family IDs for the corresponding values of the environmental
variable in env. Can be used to reorder the vector env when the order of the trios differs between env and
mat.snp.
TRUE, a 2 df likelihood ratio test is performed comparing a
gTDT model containing one indicator variable for the heterozygous genotype and one for the homozygous variant genotype with a gTDT model
additionally containing two terms for the interactions between these variables and the environmental factor. In this case, only
the arguments mat.snp, env, and famid are considered.colGxE with unstructured=FALSE, an object of class colGxE consisting of the following numeric matrices with two columns (one for each parameter):exp(coef) (see Schaid, 1996),RR,RR,model,addGandE,colGxEcolGxE with unstructured=TRUE, an object of class colGxEunstruct consisting of the following vectors:colGxEPerms,
stat,addPerms = TRUEB columns containing the values of the gTDT statistic for the SNPs when considering
the B permutations of the case-pseudo-control status,B columns containing the values of the gTDT statistic for the gene-environment interactions
when considering the B permutations of the case-pseudo-control status.model.
Schaid, D.J. (1996). General Score Tests for Associations of Genetic Markers with Disease Using Cases and Their Parents. Genetic Epidemiology, 13, 423-449.
Schwender, H., Taub, M.A., Beaty, T.H., Marazita, M.L., and Ruczinski, I. (2011). Rapid Testing of SNPs and Gene-Environment Interactions in Case-Parent Trio Data Based on Exact Analytic Parameter Estimation. Biometrics, 68, 766-773.
colTDT, ped2geno
# Load the simulated data for the analysis.
data(trio.data)
# Set up a vector with the binary environmental variable.
# Here, we consider the gene-gender interactions and
# assume that the children in the first 50 trios are
# girls, and the remaining 50 are boys.
sex <- rep(0:1, each = 50)
# Test the interaction of sex with each of the SNPs in mat.test
gxe.out <- colGxE(mat.test, sex)
# By default, an additive mode of inheritance is considered.
# If, e.g., a dominant mode should be considered, then this can
# be done by calling
gxeDom.out <- colGxE(mat.test, sex, model="dominant")
Run the code above in your browser using DataLab