R/pumaDE.R

In puma: Propagating Uncertainty in Microarray Analysis(including Affymetrix tranditional 3' arrays and exon arrays and Human Transcriptome Array 2.0)

pumaDE <- function (
	eset
,	design.matrix = createDesignMatrix(eset)
,	contrast.matrix = createContrastMatrix(eset)
)
{
	p <- matrix(0, dim(exprs(eset))[1], dim(contrast.matrix)[2]
	,	dimnames=list(rownames(exprs(eset)), colnames(contrast.matrix)))
	fold.change <- matrix(0, dim(exprs(eset))[1], dim(contrast.matrix)[2]
	,	dimnames=list(rownames(exprs(eset)), colnames(contrast.matrix)))
	for(i in 1:dim(contrast.matrix)[2])
	{
		plus_contrasts <- which(contrast.matrix[,i]==1)
		minus_contrasts <- which(contrast.matrix[,i]==-1)
		p[,i] <- pplr(
	    	cbind(exprs(eset), assayDataElement(eset,"se.exprs"))
	    ,	as.matrix(which(design.matrix[,minus_contrasts]==1,arr.ind=TRUE))[,1]
	    ,	as.matrix(which(design.matrix[,plus_contrasts]==1,arr.ind=TRUE))[,1]
		,	sorted=FALSE
	    )[,9]
		plus_arrays <- as.matrix(
			which(design.matrix[,plus_contrasts]==1,arr.ind=TRUE))[,1]
		minus_arrays <- as.matrix(
			which(design.matrix[,minus_contrasts]==1,arr.ind=TRUE))[,1]
		fold.change[,i] <-	rowMeans(log2((2^as.matrix(exprs(eset)[,plus_arrays]))+1)) -
			rowMeans(log2((2^as.matrix(exprs(eset)[,minus_arrays]))+1))
	}
	new("DEResult", statistic=p
	,	statisticDescription="Probability of Positive Log Ratio (PPLR)"
	,	FC=fold.change
	,	DEMethod="pumaDE")
}

pumaDEUnsorted <- function (
	pp
)
{
	p <- matrix(0,dim(pp[[1]])[1], dim(pp[[1]])[2])
	for(i in 1:dim(pp[[1]])[2])
	{
		sortpumaDE <- sort(pp[[2]][,i], method="quick", index.return=TRUE)[[2]]
		p[,i] <- pp[[1]][sortpumaDE,i]
	}
	return(p)
}

pplrUnsorted <- function (
	p
)
{
	sortPplr <- sort(p[,1], method="quick", index.return=TRUE)[[2]]
	pSorted <- p[,9][sortPplr]
	names(pSorted) <- rownames(p[sortPplr,])
	return(pSorted)
}