aCGH (version 1.50.0)

mergeLevels: mergeLevels

Description

Merging of predicted levels for array CGH data and similar.

Usage

mergeLevels(vecObs,vecPred,pv.thres=0.0001,ansari.sign=0.05,thresMin=0.05,thresMax=0.5,verbose=1,scale=TRUE)

Arguments

vecObs

Vector of observed values, i.e. observed log2-ratios

vecPred

Vector of predicted values, i.e. mean or median of levels predicted by segmentation algorithm

pv.thres

Significance threshold for Wilcoxon test for level merging

ansari.sign

Significance threshold for Ansari-Bradley test

thresMin

merge if segment medians are closer than thresMin , defaiult is 0.05

thresMax

don't merge if segment medians are further than thresMax (unless needs to be merged for a different reason: wilcoxon test), default is .5

verbose

if 1, progress is printed

scale

whether thresholds are on the log2ratio scale and thus need to be converted to the copy number. default is TRUE

Value

vecMerged

Vector with merged values. One merged value returned for each predicted/observed value

mnNow

Merged level medians

sq

Vector of thresholds, the function has searched through to find optimum. Note, these thresholds are based on copy number transformed values

ansari

The p-values for the ansari-bradley tests for each threshold in sq

Details

mergeLevels takes a vector of observed log2-ratios and predicted log2ratios and merges levels that are not significantly distinct.

References

Willenbrock H, Fridlyand J. (2005). A comparison study: applying segmentation to array CGH data for downstream analyses. Bioinformatics. 2005 Sep 14; [Epub ahead of print]

Examples

# Example data of observed and predicted log2-ratios
vecObs <- c(rep(0,40),rep(0.6,15),rep(0,10),rep(-0.4,20),rep(0,15))+rnorm(100,sd=0.2)
vecPred <- c(rep(median(vecObs[1:40]),40),rep(median(vecObs[41:55]),15),
  rep(median(vecObs[56:65]),10),rep(median(vecObs[66:85]),20),rep(median(vecObs[86:100]),15))

# Plot observed values (black) and predicted values (red)
plot(vecObs,pch=20)
points(vecPred,col="red",pch=20)

# Run merge function
merge.obj <- mergeLevels(vecObs,vecPred)

# Add merged values to plot
points(merge.obj$vecMerged,col="blue",pch=20)

# Examine optimum threshold
merge.obj$sq