R/normalizations.R

In rnaSeqMap: rnaSeq secondary analyses

##### all normalizations for NucleotideDistr objects
# MO, AL 11.7.11
######################################################

standarizationNormalize <- function(nd) 
{
	ndout <- nd
	dd1 <- RleList2matrix(nd@data)
	a <- dim(dd1)[1]   # num of rows
	b <- dim(dd1)[2]   # num cols
	dd <- matrix(,a,b)  
	
	for (i in 1:b){
		dd[,i]<-((dd1[,i]-mean(dd1[,i]))/sd(dd1[,i])+(abs(min(dd1[,i]))))

	}
	ndout@data <- .fillRleList(dd)
	ndout
}


# squeezes the distrib into the range 0-1

min_maxNormalize <- function(nd)
{
    ndout <- nd
	dd1 <- RleList2matrix(nd@data)
	a <- dim(dd1)[1]   # num of rows
	b <- dim(dd1)[2]   # num cols
	dd <- matrix(,a,b)  
	for (i in 1:b){
		mian <- max(dd1[,i])-min(dd1[,i])
		if (mian==0) mian <- 1			#constant coverage case, most likely 0, so we enter any constant and out is 0		 
		dd[,i]<-(dd1[,i]-min(dd1[,i]))/mian}
	ndout@data <- .fillRleList(dd)
	ndout
}

# divides each profile by the sum of coverage in this profile

densityNormalize <- function(nd)
{
	ndout <- nd
	dd1 <- RleList2matrix(nd@data)
	a <- dim(dd1)[1]   # num of rows
	b <- dim(dd1)[2]   # num cols
	dd <- matrix(,a,b) 
	
	for (i in 1:b){
		ss <- sum(dd1[,i])
		if (ss!=0) dd[,i]<- dd1[,i]/ss			
        if (ss==0) dd[,i]<- dd1[,i]     # for borderline cases when dd1[,i]==0, the output is 0 too 	
	}
	ndout@data <- .fillRleList(dd)
	ndout
}

# sums vector must have as many numbers as nd has samples

globalCountsNormalize <- function(nd, sums)
{
	
	ndout <- nd
	dd1 <- RleList2matrix(nd@data)
	a <- dim(dd1)[1]   # num of rows
	b <- dim(dd1)[2]   # num cols
	dd <- matrix(,a,b) 
	mm <- max(sums)
    
	for (i in 1:b){
			scalingFact <- mm/sums[i] 
			dd[,i]<- dd1[,i]*scalingFact}
	
	ndout@data <- .fillRleList(dd)
	ndout
}


############# utilities 
.fillRleList <- function(m) # m is a matrix, subsequent columns are changed into Rle objects
{  
	out <- list()
	cc <- dim(m)[2]
	for (i in 1:cc)
	{
		out[[i]] <- Rle(m[,i])
	}
	out
}