rareGE (version 0.1)

rareGE: Tests for gene-environment interaction and joint test of genetic main effects and gene-environment interaction

Description

Perform either gene-environment interaction tests or joint test of genetic main effects and gene-environment interaction, or all 3 tests in one function.

Usage

rareGE(phenotype, genotypes, covariates, mainweights = wuweights, 
interweights = wuweights, family = "gaussian", binomialimpute = FALSE, 
rho = seq(0, 1, by = 0.1), B = 10000, INT_FIX = TRUE, INT_RAN = TRUE, 
JOINT = TRUE)

Arguments

phenotype

a numeric vector of phenotype values.

genotypes

a matrix or a data frame for all SNPs in the test gene or genomic region. The order of rows must match the order in phenotype.

covariates

a matrix, a data frame or a vector of covariates to adjust for. The interaction between SNPs and the first column of covariates is tested. The order of rows must match the order in phenotype.

mainweights

the weight function or vector of genetic main effects (default = wuweights).

interweights

the weight function or vector of gene-environment interaction effects (default = wuweights).

family

"gaussian" for quantitative traits and "binomial" for dichotomous traits (default = "gaussian").

binomialimpute

impute missing genotypes randomly using a binomial distribution with 2 trials and success probability equal to the minor allele frequency. If FALSE, then impute missing genotypes to 0 (default = FALSE).

rho

a numeric vector with values between 0 and 1 defining the searching grid for the nuisance parameter $\rho$ in the joint test.

B

number of Monte Carlo simulations to approximate the multi-dimensional integral in calculating the p-value of the joint test (default = 10000).

INT_FIX

a logical indicator of whether the interaction test treating genetic main effects as fixed should be performed (default = TRUE).

INT_RAN

a logical indicator of whether the interaction test treating genetic main effects as random should be performed (default = TRUE).

JOINT

a logical indicator of whether the joint test of genetic main effects and gene-environment interaction effects should be performed (default = TRUE).

Value

pINT_FIX

p-value from the interaction test treating genetic main effects as fixed. NULL if INT_FIX is FALSE.

pINT_RAN

p-value from the interaction test treating genetic main effects as random. NULL if INT_RAN is FALSE.

pJOINT

p-value from the joint test of genetic main effects and gene-environment interaction effects. NULL if JOINT is FALSE.

pJOINTmin

minimum raw p-value from the searching grid of the joint test. NULL if JOINT is FALSE.

pJOINTrho

$\rho$ value where the minimum raw p-value is attained in the joint test. NULL if JOINT is FALSE.

pJOINTps

a vector of raw p-values in the joint test. NULL if JOINT is FALSE.

pJOINTinfo

a summary of the distribution of the integrand in each Monte Carlo simulation in the joint test. pJOINT is the mean excluding missing values. NULL if JOINT is FALSE.

Details

We use interaction tests and the joint test for different hypotheses. For interaction tests, genetic main effects are included in the null model, either as fixed effects (INT_FIX) or random effects (INT_RAN). We do not recommend treating genetic main effects as fixed when the number of SNPs in the gene is not small. For the joint test (JOINT), genetic main effects are not included in the null model and we are testing genetic effects, allowing for effect modification by the environmental variable. In the joint test, raw p-values are calculated for each value of the nuisance parameter $\rho$, then the minimum p-value is taken (similar to the optimal test proposed by Lee et al., 2012) and the actual p-value is calculated by multi-dimensional integration, approximated by a Monte Carlo method. This function is designed to perform two or more tests. If performing only one test, please use INT_FIX, INT_RAN or JOINT functions instead to get faster speed (only relevant models are fitted and relevant statistics are computed).

References

Chen H, Meigs JB, Dupuis J. (2014) Incorporating gene-environment interaction in testing for association with rare genetic variants. Hum Hered 78, 81-90.

Lee S, Wu MC, Lin X. (2012) Optimal tests for rare variant effects in sequencing association studies. Biostatistics 13, 762-775.

Wu MC, Lee S, Cai T, Li Y, Boehnke M, Lin X. (2011) Rare-variant association testing for sequencing data with the sequence kernel association test. Am J Hum Genet 89, 82-93.

See Also

INT_FIX, INT_RAN, JOINT

Examples

set.seed(12345)
data(rareGEgeno)
data(rareGEpheno)
# quantitative traits - testing for gene-BMI interactions
rareGE(rareGEpheno$y1, rareGEgeno, rareGEpheno[, c("bmi", "age", "sex")],
	B = 1000)
# dichotomous traits - testing for gene-BMI interactions
rareGE(rareGEpheno$y2, rareGEgeno, rareGEpheno[, c("bmi", "age", "sex")],
	family = "binomial", B = 1000)