ffpe (version 1.16.0)

CATplot: Make a Concordance at the Top plot.

Description

For the i top-ranked members of each list, concordance is defined as length(intersect(vec1[1:i],vec2[1:i]))/i. This concordance is plotted as a function of i.

Usage

CATplot(vec1, vec2, maxrank = min(length(vec1),length(vec2)), make.plot = TRUE, ...)

Arguments

vec1, vec2

Two numeric vectors, for computing concordance. If these are numeric vectors with names, the numeric values will be used for sorting and the names will be used for calculating concordance. Otherwise, they are assumed to be already-ranked vectors, and the values themselves will be used for calculating concordance.

maxrank

Optionally specify the maximum size of top-ranked items that you want to plot.

make.plot

If TRUE, the plot will be made. Set to FALSE if you just want the concordance calculations.

...

Optional arguments passed onto plot()

Value

Returns a dataframe with two columns:

i

length of top lists

concordance

fraction in common that the two provided lists have in the top i items

References

The CAT-plot was suggested by Irizarry et al.: Irizarry, R.A. et al. Multiple-laboratory comparison of microarray platforms. Nat Meth 2, 345-350 (2005).

The CAT-boxplot for multiple splits of a single dataset was suggested by Waldron et al. (under review).

Examples

library(ffpeExampleData)
data(lumibatch.GSE17565)

##preprocessing, individually for rep1 and rep2
lumibatch.rep1 <- lumibatch.GSE17565[,lumibatch.GSE17565$replicate==1]
lumbiatch.rep1 <- lumiT(lumibatch.rep1,"log2")
lumbiatch.rep1 <- lumiN(lumibatch.rep1,"quantile")
probe.var <- apply(exprs(lumibatch.rep1),1,var)
lumibatch.rep1 <- lumibatch.rep1[probe.var > median(probe.var),]

lumibatch.rep2 <- lumibatch.GSE17565[,lumibatch.GSE17565$replicate==2]
lumibatch.rep2 <- lumiT(lumibatch.rep2,"log2")
lumibatch.rep2 <- lumiN(lumibatch.rep2,"quantile")
lumibatch.rep2 <- lumibatch.rep2[featureNames(lumibatch.rep1),]

##row t-tests for differential expression
library(genefilter)
ttests.rep1 <- rowttests(exprs(lumibatch.rep1),fac=factor(lumibatch.rep1$cell.type))
ttests.rep2 <- rowttests(exprs(lumibatch.rep2),fac=factor(lumibatch.rep2$cell.type))

pvals.rep1 <- ttests.rep1$p.value;names(pvals.rep1) <- rownames(ttests.rep1)
pvals.rep2 <- ttests.rep2$p.value;names(pvals.rep2) <- rownames(ttests.rep2)

## Very high concordance between top differentially expressed gene lists
## identified by different replicates
x <- CATplot(pvals.rep1,pvals.rep2,maxrank=1000,xlab="Size of top-ranked gene lists",ylab="Concordance")
legend("topleft",lty=1:2,legend=c("Actual concordance","Concordance expected by chance"), bty="n")