pathWave {PathWave}R Documentation

PathWave analysis tool

Description

Performs an enrichment analysis on optimally arranged grids. Features are generated using a Haar wavelet transform.

Usage

pathWave(x, y, optimalM,  mTest = TRUE, mTestMethod = "Bonferroni", pvalCutoff = 0.01, genes=NULL, diffReac = 5, nperm = 10000, verbose = TRUE)

Arguments

x Matrix of expression values. Names of row elements correspond to elements in optimalM. Columns are data samples.
y Class factor for x. Should consist of only two classes and length(y) should correspond to ncol(x).
optimalM List of optimal arranged grids as returned by function pwOptGrids(). See details.
mTest Should the results be corrected for multiple testing (default TRUE)?
mTestMethod Method for multiple testing correction (default "Bonferroni") as defined by package multtest.
pvalCutoff Significance level (default 0.01).
genes List of genes for each reaction per pathway as returned by pwKEGGxml(). See details.
diffReac Pathways with how many differentially expressed reaction should be considered (default 5)?
nperm Number of permutations that should be used to estimate the underlying distribution (default 10,000)?
verbose Should the progress in permutations be printed after every 100 permutations?

Details

x
Rownames(x) should correspond to the reactionIDs in optimalM as the values of x will be mapped onto the matrices. If for an entry in optimalM no entry in x can be found, it is set to 0.
optimalM
List of the optimal arranged grid for each pathway. The Matrix consists of "0" and the Reaction IDs (see exampoe below). The structure of the list: optimalM$pathwayID$M .
genes
List element "genes" from list returned by pwKEGGXml(). For every pathwayID a vector is stored with the geneIDs. The names of the vector are the reactionIDs. The reactionIDs can occur more than once, if more than one gene is mapped onto the reaction.

Value

PathWave returns a list of class PathWave:

pathwayID The ID of enriched pathways.
p.value P-value of the enriched pathway. Correspond to the p-value of the highest ranked feature.
score Size of the feature with which the score was calculated. This is a measurement of the size of the significant pattern.
reactions Reactions from which the significant features were calculated by a Haar wavelet transform.
reaction.p.value The p-values of the reactions. Calculated by a t-test on x.
reaction.regulation The regulation of the reactions: +1 up-regulated, -1 down-regulated, 0 not differentially regulated in class levels(y)[1].
feature.p.value The p-values of the features.
feature List of the significant features.


In List features: Each sub-matrix is listed from which the significant features were derived.

Author(s)

Gunnar Schramm

References

iChip

See Also

See Also pwOptGrids,pwKEGGxml

Examples

#build test optimal matrices:
test.opM=list()
test.opM[["test1"]]=list(M=matrix(sample(c(paste("test1:R",1:6,sep=""),rep("0",19))),nrow=5))
test.opM[["test2"]]=list(M=matrix(sample(c(paste("test2:R",7:11,sep=""),rep("0",11))),nrow=4))

#generate test class factor for two classes
test.y=as.factor(c(rep("test.1",20),rep("test.2",20)))

#generate test expression matrix with two classes
test.x1=matrix(rnorm(11*length(test.y)),ncol=length(test.y),dimnames=list(paste("test1:R",1:11,sep=""),paste("sample",1:length(test.y),sep="")))
test.x2=matrix(rnorm(11*length(test.y)),ncol=length(test.y),dimnames=list(paste("test2:R",1:11,sep=""),paste("sample",1:length(test.y),sep="")))

#all reactions of the second class are up-regulated in the first pathway
test.x1[,test.y==levels(test.y)[2]]=test.x1[,test.y==levels(test.y)[2]] + 2

test.x=rbind(test.x1,test.x2)

#build gene list
test.gene=list()
test1.genes=as.character(1:6)
test2.genes=as.character(1:5)
names(test1.genes)=paste("test1:R",1:6,sep="")
names(test2.genes)=paste("test2:R",7:11,sep="")
test.gene[["genes"]]=list(test1=test1.genes,test2=test2.genes)

#start pathwave
test.result=pathWave(test.x,test.y,optimalM=test.opM,pvalCutoff=0.05,genes=test.gene$genes,diffReac=1)

[Package PathWave version 1.0 Index]