genefu (version 2.4.2)

compare.proto.cor: Function to statistically compare correlation to prototypes

Description

This function performs a statistical comparison of the correlation coefficients as computed between each probe and prototype.

Usage

compare.proto.cor(gene.cor, proto.cor, nn,
  p.adjust.m = c("none", "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr"))

Arguments

gene.cor

Correlation coefficients between the probes and each of the prototypes.

proto.cor

Pairwise correlation coefficients of the prototypes.

nn

Number of samples used to compute the correlation coefficients between the probes and each of the prototypes.

p.adjust.m

Correction method as defined in p.adjust.

Value

• Data frame with probes in rows and with three columns: "proto" is the prototype to which the probe is the most correlated, "cor" is the actual correlation, and "signif" is the (corrected) p-value for the superiority of the correlation to this prototype compared to the second highest correlation.

See Also

compute.proto.cor.meta, compute.pairw.cor.meta

Examples

## load VDX dataset
data(vdxs)
## load NKI dataset
data(nkis)
## reduce datasets
ginter <- intersect(annot.vdxs[ ,"EntrezGene.ID"], annot.nkis[ ,"EntrezGene.ID"])
ginter <- ginter[!is.na(ginter)][1:30]
myx <- unique(c(match(ginter, annot.vdxs[ ,"EntrezGene.ID"]),
  sample(x=1:nrow(annot.vdxs), size=20)))
data2.vdxs <- data.vdxs[ ,myx]
annot2.vdxs <- annot.vdxs[myx, ]
myx <- unique(c(match(ginter, annot.nkis[ ,"EntrezGene.ID"]),
  sample(x=1:nrow(annot.nkis), size=20)))
data2.nkis <- data.nkis[ ,myx]
annot2.nkis <- annot.nkis[myx, ]
## mapping of datasets
datas <- list("VDX"=data2.vdxs,"NKI"=data2.nkis)
annots <- list("VDX"=annot2.vdxs, "NKI"=annot2.nkis)
datas.mapped <- map.datasets(datas=datas, annots=annots, do.mapping=TRUE)
## define some prototypes
protos <- paste("geneid", ginter[1:3], sep=".")
## compute meta-estimate of correlation coefficients to the three prototype genes
probecor <- compute.proto.cor.meta(datas=datas.mapped$datas, proto=protos,
  method="pearson")
## compute meta-estimate of pairwise correlation coefficients between prototypes
datas.proto <- lapply(X=datas.mapped$datas, FUN=function(x, p) {
  return(x[ ,p,drop=FALSE]) }, p=protos)
protocor <- compute.pairw.cor.meta(datas=datas.proto, method="pearson")
## compare correlation coefficients to each prototype
res <- compare.proto.cor(gene.cor=probecor$cor, proto.cor=protocor$cor,
  nn=probecor$cor.n, p.adjust.m="fdr")
head(res)