GSAgm (version 1.0)

RNAgamma: RNA gamma

Description

For GSA of RNA-seq data, the following procedure, similar to the analysis of SNP data, is implemented (see Fridley et al[2] for more details on the method). Step 1: Association of gene expression data from RNA-seq (count data) is assessed for differential expression between two groups using edgeR[3]. Step 2: P-values from the association analysis within edgeR for genes within a given gene set are combined using the Gamma Method to determine the association of the gene set with the phenotype. Currently, the RNA-seq GSA allows only a binary phenotype (i.e, treatment, control).

Usage

RNAgamma(formula, data, rnaprefix="ENSG", gammaShape=1,STT=NULL, pheno.type=c("case.control"),tagwise=F,perm=T,n.perm=1000,seed=12212012)

Arguments

formula

formula in R format: phenotype~cov1+cov2

data

data frame containing phenotype, covars, and RNA stuff

rnaprefix

RNA data prefix, defaults to ENSG ensembl genes

gammaShape

numeric indicating the gamma shape parameter to be used for p-value summarization

STT

numeric indicating soft truncation threshold to be used, will calculate gamma parameter (must be

pheno.type

type of phenotype, case-control results in logistic regression, quantitative results in OLS, and survival results in cox model

tagwise

TRUE or FALSE for estimating tagwise dispersion values by an empirical Bayes method based on weighted conditional maximum likelihood. Defaults to maximizing the negative binomial conditional common likelihood for the common dispersion across all tags.

perm

boolean indicating whether permutation p-value are to be used for the gamma summary method

n.perm

numeric indicating number of permutations to be used

seed

numeric to set RNG for reproducability

Examples

data(testdata)
data(rnaseq_counts)
testdata <- cbind(testdata,rnaseq_counts)
RNAgamma(pheno~strata(study)+age, data=testdata, rnaprefix="rnaseqcount", 
         pheno.type=c("case.control"),tagwise=FALSE,perm=TRUE,n.perm=5)

##No covars, no permutation
RNAgamma(pheno~., data=testdata, rnaprefix="rnaseqcount", 
         pheno.type=c("case.control"),tagwise=FALSE,perm=FALSE)