RP.symetric: RP symetric & multicore

In galder-max/RPlite: What the package does (short line)

Usage

RP.symetric(lMs, lOs, nperms = 100, size = 64, mc.cores = NA)

Arguments

lMs	list of matrices in the meta-analysis. A list containing one matrix can be passed. Or a list containing multiple matrices, with same number of rows and same ordering of rows (the mapping is not performed, nor checked for!)
lOs	list of 2-level factors corresponding to the matrices in lMs. Attention must be paid to the order of the levels in each factor. Ratios are always level1/level2. Ideally, each vector should heave the two same unordered levels.
nperms	number of permutations to perform. 100 is typically good. Whereas RP follows a f(gamma distribution) for paired expriments, it is not the case for unpaired experiments. The null distribution (ND) is approximated by permutations. The more permutations, the better the approximation, the more time it takes to compute the ND.
size	Very useful. this tells the function to cut the matrix of possible ratios is smaller matrices with size columns. It reduces memory consumption by very little increase in time consumption. Memory usage can get critical, especially for big meta-analyses.
mc.cores	Number of cores to use if parallelized. Decrease time consumption proportionnaly but increases memory needs.

Author(s)

[e-mail]

Examples

## samples
nc1<-20
nc2<-10
##genes
nr<-2000
## matrix of expression
m1<-matrix(rnorm(nc1*nr),nr,nc1)
m2<-matrix(rnorm(nc2*nr),nr,nc2)

##factor of paired leves
o1<-as.factor(c(rep("level1",nc1/2),rep("level2",nc1/2)))
o2<-as.factor(c(rep("level1",nc2/2),rep("level2",nc2/2)))

## Rank Product for first matrix
resRP<-RP.symetric(list(m1),list(o1),nperms=100)
head(resRP$lStats)


##Down regulated Genes
sum(resRP$lStats$qvals.do<0.05)
##Up regulated Genes
sum(resRP$lStats$qvals.do<0.05)




## meta-analysis with Rank Product (2 datasets/matrices) with same
## number of rows and ordered rows (mapping is not performed)!
resRPmeta<-RP.symetric(list(m1,m2),list(o1,o2),nperms=100)

head(resRPmeta$finalRes)
##Down regulated Genes
sum(resRPmeta$lfinalRes$qvals.do<0.05)
##Up regulated Genes
sum(resRPmeta$lfinalRes$qvals.up<0.05)



## ADDING signal (mean fold change=2,sd=.1) to 50 first genes
wsignal<-1:50
m1[wsignal,1:(nc1/2)]<-m1[wsignal,1:(nc1/2)]-rnorm(length(m1[wsignal,1:(nc1/2)]),mean=2,sd=.1)
m2[wsignal,1:(nc2/2)]<-m2[wsignal,1:(nc2/2)]-rnorm(length(m2[wsignal,1:(nc2/2)]),mean=2,sd=.1)


resRPmeta<-RP.symetric(list(m1,m2),list(o1,o2),nperms=100)
##Down regulated Genes
sum(resRPmeta$finalRes$qvals.do<0.05)
##Up regulated Genes
sum(resRPmeta$finalRes$qvals.up<0.05)


## plot scores for all genes
cols<-rep("grey",nr)
cols[resRPmeta$finalRes$qvals.up<0.05]<-"red"
barplot(resRPmeta$finalRes$scores,
border=NA,
col=cols,
ylim=c(-500,500),
ylab="RP scores")
## add thresholds of RP scores corresponding to bonferroni-corrected
## p-values 0.05
abline(h=quantile(unlist(resRPmeta$null.distrib),
probs=c(0.05/nr,1-0.05/nr)),
col="grey",lty=2)

galder-max/RPlite documentation built on May 5, 2019, 3:49 a.m.