[BioC] suggestions for plotPCA / pca.legend

[Jenny Drnevich]

Hi Jim,

I've been playing around with the versatility of plotPCA - thanks for 
adding support for matrices! Say I have a 2x2 factorial design, and I want 
to plot the 4 groups, but not using 4 separate symbols, each with its own 
color. Instead, I want use 2 symbols to indicate the levels of the first 
factor and 2 colors to indicate the levels of the second factor. I can do 
this by using the arguments "pch" and "col" instead of "groups". However, 
the legend doesn't come out correctly - instead of the 4 symbol/color 
groups, there are only 2 - one symbol in one color and the other symbol in 
the other color. I looked at pca.legend and I think I've come up with a 
fix. At the very beginning, instead of:

pch <- sort(unique(pch))
col <- sort(unique(col))

this worked to give the correct legend:

gps <- sort(unique(paste(pch,col,sep="A")))
pch <- as.numeric(sapply(strsplit(gps,"A"),function(x) x[1]))
col <- as.numeric(sapply(strsplit(gps,"A"),function(x) x[2]))


It might also be good to include some sort of warning if 
length(groupnames)!=length(gps); if you forget to put in groupnames in the 
original call to plotPCA, it defaults to the names of the columns, and the 
legend symbols get recycled so that the legend is incorrect.

Cheers,
Jenny





Jenny Drnevich, Ph.D.

Functional Genomics Bioinformatics Specialist
W.M. Keck Center for Comparative and Functional Genomics
Roy J. Carver Biotechnology Center
University of Illinois, Urbana-Champaign

330 ERML
1201 W. Gregory Dr.
Urbana, IL 61801
USA

ph: 217-244-7355
fax: 217-265-5066
e-mail: drnevich at uiuc.edu