#plot a dot per __row__ in table
args <- commandArgs(trailingOnly = TRUE)
intab <- args[1]

data=read.table(intab, sep="\t", row.names=1, header=TRUE)
#remove any constant columns (at least 1 value must be different
data=data[,apply(data, 2, var, na.rm=TRUE) != 0]
pca=prcomp(data, scale=TRUE)
library(ggplot2)
scores=data.frame(pca$x)
cols=ncol(scores)
cells=gsub("_.*", "", rownames(scores))
library(ggplot2)
scores=data.frame(scores,cells)
colnames(scores)[cols+1]="cell"
#palette Set1 when 9 or less, Paired for up to 12
ggplot(scores, aes(x=PC1, y=PC2))+geom_point(aes(color=cell), size=6)+geom_text(aes(label=rownames(scores), hjust=-.2), size=3)+scale_colour_brewer(palette="Set1")

ggplot(scores, aes(x=PC2, y=PC3))+geom_point(aes(color=cell), size=6)+geom_text(aes(label=rownames(scores), hjust=-.2), size=3)+scale_colour_brewer(palette="Set1")

ggplot(scores, aes(x=PC3, y=PC4))+geom_point(aes(color=cell), size=6)+geom_text(aes(label=rownames(scores), hjust=-.2), size=3)+scale_colour_brewer(palette="Set1")

#variance plot
#http://strata.uga.edu/6370/lecturenotes/principalComponents.html
sd=pca$sdev
var=sd^2
var.percent=var/sum(var) * 100
barplot(var.percent, xlab="PC", ylab="Percent Variance", names.arg=1:length(var.percent), las=1, ylim=c(0,max(var.percent)), col="blue")
abline(h=1/ncol(pca)*100, col="red")