Indi.DE.Analysis: Main Function for Individual Study DE: microarray & RNAseq.

In metaOmic/MetaDE: MetaDE: Transcriptome meta-analysis for differentially expressed gene detection

Description

Main Function for Individual Study DE: microarray & RNAseq The Indi.DE.Analysis is a function to perform individual association analysis between gene expression and the response/phenoype of interest (can be either group, continuous or survival).

Usage

Indi.DE.Analysis(data, clin.data, data.type, resp.type, response,
  covariate = NULL, ind.method, select.group = NULL, ref.level = NULL,
  paired = NULL, asymptotic = NULL, nperm = NULL, tail = "abs",
  seed = 12345, ...)

Arguments

data	is a list of K elements, where K is the number of studies, each element is a microarray or RNAseq expression matrix with G rows and N columns, where G is number of matched genes and N is the sample size.
clin.data	is a list of K elements, each element includes is a clinical data frame with N rows and p columns, where N is the sample size and p is the number of clinical variables (main response included).
data.type	is a character indicating the data type of the elements in data, must be "continuous" or "discrete".
resp.type	is a character indicating the response type of the response variable selected, must be one of "twoclass", "multiclass", "continuous" and "survival".
response	is one column name of clin.data, indicating the phenotype of interest. For survival, two column names have to be specified, the first is the survival time and the second is the censoring status.
covariate	are the clinical covariates you wish to adjust for in the DE analysis, can be a vector of column names or NULL.
ind.method	is a character vector to specify the method used to test if there is association between the gene expression and outcome variable. must be one of "limma", "sam" for "continuous" data type and "edgeR", "DESeq2" or "limmaVoom" for "discrete" data type.
select.group:	for two-class comparison only, specify the two groups for comparison when the group factor has more than two levels.
ref.level:	for two-class/multi-class comparison only, specify the reference level of the group factor.
paired:	logical value indicating whether paired design;
asymptotic:	a logical value indicating whether asymptotic distribution should be used. If FALSE, permutation will be performed.
nperm:	the number of permutations. Applicable when asymptotic is FALSE.
tail:	a character string specifying the alternative hypothesis, must be one of "abs" (default), "low" or "high". For resp.type = "continuous", "survival" only.
seed:	Optional initial seed for random number generator.

Value

a list with components:

• p: For all types of response, the p-value of the association test for each gene

• stat: For "continuous" and "survival" only, the value of test statistic for each ##' gene

• bp: For "continuous" and "survival" only, the p-value from nperm: permutations for each gene. It will be used for the meta analysis by default. It can be NULL if you chose asymptotic results.

• log2FC: For "twoclass" only, the log2 fold change for each gene

• lfcSE: For "twoclass" only, the standard error of log2 fold change for each gene

Examples

data('Leukemia')
data('LeukemiaLabel')
data <- Leukemia
K <- length(data)
clin.data <- lapply(label, function(x) {data.frame(x)} )
for (k in 1:length(clin.data)){
 colnames(clin.data[[k]]) <- "label"
}
select.group <- c('inv(16)','t(15;17)')
ref.level <- "inv(16)"
data.type <- "continuous"
ind.method <- c('limma','limma','limma')
resp.type <- "twoclass"
paired <- rep(FALSE,length(data))
ind.res <- Indi.DE.Analysis(data=data,clin.data= clin.data, 
                        data.type=data.type,resp.type = resp.type,
                        response='label',
                        ind.method=ind.method,select.group = select.group,
                        ref.level=ref.level,paired=paired)
N <- sapply(data, FUN=function(x) ncol(x))
survival.time <- lapply(N,FUN = function(x) round(runif(x,10,2000)))
censor.status <- lapply(N,FUN = function(x) sample(c(0,1),x,replace=TRUE) )
for (k in 1:length(clin.data)){
  clin.data[[k]] <- cbind(clin.data[[k]],survival.time[[k]],censor.status[[k]])
  colnames(clin.data[[k]])[2:3] <- c("survival","censor")
}
ind.method <- c('logrank','logrank','logrank')
resp.type <- "survival"
ind.res <- Indi.DE.Analysis(data=data,clin.data= clin.data, 
                         data.type=data.type,resp.type = resp.type,
                         response=c("survival","censor"),
                         ind.method=ind.method,asymptotic=TRUE)

metaOmic/MetaDE documentation built on May 22, 2019, 6:54 p.m.