Data visulizations: PCA, scatter plot, heatmap, cloud plot

September 13, 2013, 12:42:55 PM

According to literatures, there are so many kinds of plot to present LC/MS-derived metabolomic data such as principal component analysis (PCA), scatter plot, heatmap, volcano plot. As a green hand of xcms, I don't know how to use it for generating these plots. In addition, xcms online can present data in Cloud Plot. I wanna know whether it is possible to produce Cloud plot using xcms. If so, how to generate using xcms? I am desperately waiting for the answers for the questions. Your prompt responses are greatly appreciated in advance.

Best wishes,

Yong

Re: Data visulizations: PCA, scatter plot, heatmap, cloud pl

Reply #1 – September 15, 2013, 03:09:14 PM

Yong,

You might want to check out the cookbook and FAQ's. We have put a lot of code for sorting different problems and frequent requests.
PCA: viewtopic.php?f=26&t=124
PLS: viewtopic.php?f=26&t=123
Heatmaps: viewtopic.php?f=26&t=385&p=1233#p1233
As for the cloud plot, it came out along side online xcms and has not been included into xcms.
I've put a little code together below for something similar however, things like the m/z scale of the points, m/z y-axis (currently showing intensity) and continuous RT plot could probably use some playing with to make it more suitable. For the RT plot I just used a mean Intensity across all samples. I did this pretty quickly, so improvements could be easyly found. You could also wrap it into the playwith package or the googleviz package to use for identification of which point is which

Code: [Select]

makePseudoCloud<-function(fill, classA, classB, pval){
	require(xcms)
	diff<-diffreport(fill, classA, classB, eicmax=0, file=NULL)

	gr<-groups(fill)
	d<-ncol(gr)
	rtmed<-gr[,"rtmed"]
	

	dat<-data.frame(X=c(rtmed, rtmed), Y=c(rowMeans(diff[,(d+6):ncol(diff)]), -(rowMeans(diff[,(d+6):ncol(diff)])) ))
	plot(dat, type="h")
	abline(0,0) ## make the plot
	
	ix<-which(diff[,"pvalue"] < pval)
	fold<-max(diff[ix,"fold"])
	m.mz<-max(diff[,"mzmed"])
	m.int<-max(dat[,"Y"])

	for(i in ix){
		if(diff[i, "tstat"] > 0){
			foo<-(diff[i,"mzmed"]/m.mz)*m.int
			#foo<-mean(as.numeric(diff[i,(d+6):ncol(diff)]))
			points(diff[i,"rtmed"], foo, col="blue", pch=16, cex=5*(diff[i,"fold"]/max(fold)) )
		} else if(diff[i, "tstat"] < 0){
			foo<-(diff[i,"mzmed"]/m.mz)*m.int
			points(diff[i,"rtmed"], -foo, col="red", pch=16, cex=5*(diff[i, "fold"]/max(fold) ))
		}
	}
}