maskBAD (version 1.12.0)

overlapExprExtMasks: Error Analysis of Masking Results

Description

Expression mask results for a range of cutoff values are compared with an external mask (for example a mask based on sequence data) and type 1 and type 2 errors are estimated.

Usage

overlapExprExtMasks(probes,seqdata,cutoffs="none",wilcox.ks=FALSE,sample=10,plotCutoffs=TRUE,verbose=TRUE)

Arguments

probes

A matrix with 3 columns. The first and second column represent the x and y coordinates on the Microarray. The third column contains a quality entry for each probe, e.g. the quality score obtained from mask analysis.

seqdata

A matrix with 3 columns containing x, y coordinates and 0,1 entries in column 3, defining whether a probe has a sequence difference (0) or not.

cutoffs

A vector including all cutoff values for the quality scores of an expression mask that should be used for the error analysis. If no cutoffs are given (default is "none") the cutoffs are the quantiles of the quality scores starting from 0 to 1 in steps of 0.01.

wilcox.ks

Logical, default=FALSE element determining whether the Kolmogorow-Smirnow Test and Wilcoxon Rank Test analysis should be performed (see reference below).

sample

To compare the p value distribution with the Kolmogorow-Smirnow Test and Wilcoxon Rank Test for different cutoffs, the sampling option can be used to compute the quality score distribution for different cutoffs. This value indicates how often the sampling should be performed.

plotCutoffs

Logical, default=TRUE element determining whether the cutoffs should be drawn in the overlap plot.

verbose

Logical. If 'TRUE', it writes out some messages indicating progress. If 'FALSE' nothing should be printed.

Value

A list of five objects will be returned.

type1

A vector of the type 1 error for each cutoff.

type2

A vector of the type 2 error for each cutoff.

confT1

A matrix with the upper (column 1) and lower (column 2) confidence intervals for the type 1 error.

confT2

A matrix with the upper (column 1) and lower (column 2) confidence intervals for the type 2 error.

ksP

If wilcox.ks is 'TRUE', a vector of quality scores from a two sample Kolmogorov-Smirnov comparing distributions of quality score for probes designated as BAD and not in external mask.

wilcoxonP

If wilcox.ks is 'TRUE', a vector of quality scores from a two sample Wilcoxon rank test comparing distributions of quality score for probes designated as BAD and not in external mask.

ksBoot

For each cutoff sample(default=10) times cutoff values for the Kolmogorov-Smirnov test will be generated.

wilcoxBoot

For each cutoff sample(default=10) times cutoff values for the wilcoxon rank sum test will be generated.

cutoffs

List of cutoffs used for the error analysis

testCutoffs

If wilcox.ks is 'TRUE', a list with cutoff information will be provided. The first list entry includes all cutoffs used in the two sample Kolmogorov-Smirnov test and the two sample wilcoxon rank sum test analysis will be produced. A cutoff can appear sample(default=10) times. In theory there should be sample times the number of cutoff values entries in this vector, but usually there are fewer entries, because for certain cutoff values, it is not possible to calculate the exact p value in one of the tests. The second list entry transforms the cutoffs in ranks and can be used for the plotting of the test results.

Details

The function overlapExprExtMasks compares expression mask results with an external (for example sequence-based) mask and might help to choose a quality score cutoff for masking probes.

References

Dannemann et al, The effects of probe binding affinity differences on gene expression measurements and how to deal with them. Bioinformatics 2009

See Also

mask, prepareMaskedAffybatch, plotProbe

Examples

## loading mask on all genes (exmask1) of the same dataset
data(exmask)
overlapExSeq <- overlapExprExtMasks(exmask$probes[,1:3],sequenceMask[,c(1,2,4)])

## plot results
plot(overlapExSeq$type1,overlapExSeq$type2,type="l",col="red",
     main="Overlap expression based mask - sequence based mask",xlab="Type 1",ylab="Type 2")
abline(1,-1,col="gray")

## performing wilcoxon rank sum test and Kolmogorov-Smirnov test on
## expression mask with all genes (exmask)
overlapTests <-
  overlapExprExtMasks(exmask$probes[,1:3],sequenceMask[,c(1,2,4)],wilcox.ks=TRUE)
layout(matrix(1:2,ncol=1))
plot(overlapTests$testCutoff[[1]],overlapTests$ksBoot,col="red",main="Kolmogorov-Smirnov Test",xlab="Quality score cutoff",
     ylab="p value (Kolmogorov-Smirnov Test)",ylim=c(0,1),pch=16,xaxt="n")
axis(1,at=1:length(unique(overlapTests$testCutoff[[2]])),labels=signif(unique(overlapTests$testCutoff[[2]]),2),las=3)
lines(which(unique(overlapTests$testCutoff[[2]]) %in% overlapTests$testCutoff[[2]]),overlapTests$ksP[!is.na(overlapTests$ksP)],type="p",pch=16,cex=0.8)
plot(overlapTests$testCutoff[[1]],overlapTests$wilcoxonBoot,col="green",main="Wilcoxon Rank Sum Test",xlab="Quality score cutoff",
     ylab="p value (Wilcoxon Rank Sum Test)",ylim=c(0,1),pch=16,xaxt="n")
axis(1,at=1:length(unique(overlapTests$testCutoff[[2]])),labels=signif(unique(overlapTests$testCutoff[[2]]),2),las=3)
lines(which(unique(overlapTests$testCutoff[[2]]) %in% overlapTests$testCutoff[[2]]),overlapTests$wilcoxonP[!is.na(overlapTests$wilcoxonP)],type="p",pch=16,cex=0.8)