parSegment {ParDNAcopy}

Parallel implementation of segment function of DNAcopy

Description

There are three key differences between this function and the original segment function of package DNAcopy. First, the execution can be parallelized, either by using multiple cores of the present host or by invoking a grid engine to run on multiple hosts. Secondly, random number generator may be re-initialized, with the same seed, for each sample. Finally, there is a "skinny" option for the value, i.e., a DNAcopy object with no data item.

Usage

parSegment(CNAobj, ranseed = NULL, distrib = c("vanilla", "Rparallel"), 
njobs = 1, out = c("full", "skinny"), ...)

Arguments

CNAobj

An object of class CNA, usually a value produced by the CNA function of DNAcopy

ranseed

A single integer to seed the random number generator.

distrib

One of "vanilla" (default) and "Rparallel" to choose a parallelization option: no parallelization ("vanilla"), parallelization on multiple cores of the local host ("Rparallel").

njobs

An integer specifying the desired number of parallel jobs.

out

One of "full" (default) or "skinny" to specify the form of the value, an object of class DNAcopy, with the data item present ("full") or not ("skinny").

...

Arguments other than x to be passed on to the segment function of DNAcopy.

Value

An object of class DNAcopy. If out == "skinny" the data item of the value will not be returned in order to reduce the memory use.

Author(s)

Alex Krasnitz

See Also

Package DNAcopy.

Examples

data(coriell)
#prepare data for segmentation
CNA.object <- CNA(genomdat=coriell[,c(4,5)],coriell$Chromosome,coriell$Position,
	data.type="logratio",sampleid=dimnames(coriell)[[2]][4:5])
#equivalent to "segment" of DNAcopy
parseg<-parSegment(CNA.object,undo.splits="sdundo")
#Random number generator to be re-seeded for each sample
parsegrep<-parSegment(CNA.object,ranseed=123,undo.splits="sdundo")
#multi-core execution but the result should not change
parsegrep1<-parSegment(CNA.object,ranseed=123,distrib="Rparallel",njobs=2,
undo.splits="sdundo")