cd02-0-plotColoredClusters: Plot Dendrograms with Color-Coded Labels

In ClassDiscovery: Classes and Methods for "Class Discovery" with Microarrays or Proteomics

Plot Dendrograms with Color-Coded Labels

Description

Provides an interface to the plot method for hclust that makes it easier to plot dendrograms with labels that are color-coded, usually to indicate the different levels of a factor.

Usage

plotColoredClusters(hd, labs, cols, cex = 0.8, main = "", line = 0, ...)
pcc(hd, colors=NULL, ...)

Arguments

hd	An object with S3 class hclust, as produced by the hclust function.
labs	A vector of character strings used to label the leaves in the dendrogram

.

cols	A vector of color names suitable for passing to the col argument of graphics routines.
cex	A numeric value; the character expansion parameter of par.
main	A character string; the plot title
line	An integer determining how far away to plot the labels; see mtext for details.
colors	A list; see details.
...	Any additional graphical parameters that can be supplied when plotting an hclust object.

Details

The plotColoredClusters function is used to implement the pltree methods of the Mosaic class and the PCanova class. It simply bundles a two step process (first plotting the dendrogram with no labels, followed by writing the labels in the right places with the desired colors) into a single unit.

The pcc function also produces dendrograms with colored annotations. However, instead of coloring the labels based on a single factor, it produces color bars for any number of factors. The colors argument should be a list with named components, where each component should correspond to a factor and a color scheme. Specifically, the components must themselves be lists with two components named fac (and containing the factor) and col (containing a named vector specifying colors for each level of the factor).

Value

The function has no useful return value; it merely produces a plot.

See Also

hclust, Mosaic, PCanova, par

Examples

# simulate data from three different groups
d1 <- matrix(rnorm(100*10, rnorm(100, 0.5)), nrow=100, ncol=10, byrow=FALSE)
d2 <- matrix(rnorm(100*10, rnorm(100, 0.5)), nrow=100, ncol=10, byrow=FALSE)
d3 <- matrix(rnorm(100*10, rnorm(100, 0.5)), nrow=100, ncol=10, byrow=FALSE)
dd <- cbind(d1, d2, d3)

# perform hierarchical clustering using correlation
hc <- hclust(distanceMatrix(dd, 'pearson'), method='average')
cols <- rep(c('red', 'green', 'blue'), each=10)
labs <- paste('X', 1:30, sep='')

# plot the dendrogram with color-coded groups
plotColoredClusters(hc, labs=labs, cols=cols)

# simulate another dataset
fakedata <- matrix(rnorm(200*30), ncol=30)
colnames(fakedata) <- paste("P", 1:30, sep='')
# define two basic factors, with colors
faccol <- list(fac=factor(rep(c("A", "B"), each=15)),
               col=c(A='red', B='green'))
fac2col <- list(fac=factor(rep(c("X", "Y", "Z"), times=10)),
               col=c(X='cyan', Y='orange', Z='purple'))
# add another factor that reverses the colors
BA <- faccol
BA$col <- c(A='blue', B='yellow')
# assemble the list of factors
colors <- list(AB=faccol, XYZ=fac2col, "tricky long name"=fac2col,
               another=BA)
# cluster the samples
hc <- hclust(distanceMatrix(fakedata, "pearson"), "ward")
# plot the results
pcc(hc, colors)


#cleanup
rm(d1, d2, d3, dd, hc, cols, labs)
rm(fakedata, faccol, fac2col, BA, colors)

ClassDiscovery documentation built on Sept. 2, 2024, 3 a.m.