pasta {subgxe}

pasta for multi-phenotype analysis

Description

Search for the subset that yields the strongest evidence of association and calculate the meta-analytic p-value, possibly in the presence of gene-environmental interaction.

Usage

pasta(p.values, study.sizes, cor)

Arguments

p.values

The p.value of each study.

study.sizes

The sample size of each study.

cor

The correlation matrix of the studies. For example, if each study is independent, cor would be the identity matrix.

Value

A list containing the joint p value and the test statistic, which contains the optimal subset.

References

Yu Y, Xia L, Lee S, Zhou X, Stringham H, M, Boehnke M, Mukherjee B: Subset-Based Analysis Using Gene-Environment Interactions for Discovery of Genetic Associations across Multiple Studies or Phenotypes. Hum Hered 2019. doi: 10.1159/000496867

Examples

# grab synthetic study for example
data("studies")
n.studies <- 5
study.sizes <- c(nrow(studies[[1]]), nrow(studies[[2]]), nrow(studies[[3]]),
                   nrow(studies[[4]]), nrow(studies[[5]]))
study.pvals <- rep(0, n.studies)
# Correlations of p-values among the studies.
# In this case the studies were generated independently so its just I
cor.matrix <- diag(1, n.studies)
# load the lrtest() function to conduct the likelihood ratio test
# Used just to generate the input p-values, not required in pasta itself.

library(lmtest)

for(i in 1:n.studies) {
 # model with gene(G) by environment(E) interaction
 model <- glm(D ~ G + E + GbyE, data = studies[[i]], family = binomial)
 # model without G and GE interaction
 null.model <- glm(D ~ E, data = studies[[i]], family = binomial)
 # likelihood ratio test from the package lmtest
 study.pvals[i] = lmtest::lrtest(null.model, model)[2, 5]
}

pasta <- pasta(study.pvals, study.sizes, cor.matrix)

pasta$p.pasta
pasta$test.statistic$selected.subset