Loading plot with R

Question

[Giffredo]

Hi,

I have a problem concerning R code. I looking for a method to reach a loading plot. I made a scatter plot of my data. I used this code so far:

library(labdsv)
PCO.1<-pco(data.dist, k=2)
PCO.1
plot.pco(PCO.1)

I hope I gave all the information. Thank you in advance,

Giffredo

[komal.rathi]

How did you get data.dist? What is the problem/error?

Did you try using just plot(PCO.1) instead of plot.pco(PCO.1)?

[Giffredo]

Hi, thanks for the reply!

This is the code used to reach the data.dist:

data=read.table("genere.txt", header=T, row.names=1, dec=".", sep="\t")
> dist.JSD <- function(inMatrix, pseudocount=0.000001, ...) {
+   KLD <- function(x,y) sum(x *log(x/y))
+   JSD<- function(x,y) sqrt(0.5 * KLD(x, (x+y)/2) + 0.5 * KLD(y, (x+y)/2))
+   matrixColSize <- length(colnames(inMatrix))
+   matrixRowSize <- length(rownames(inMatrix))
+   colnames <- colnames(inMatrix)
+   resultsMatrix <- matrix(0, matrixColSize, matrixColSize)
+   inMatrix = apply(inMatrix,1:2,function(x) ifelse (x==0,pseudocount,x))
+   for(i in 1:matrixColSize) {
+     for(j in 1:matrixColSize) {
+       resultsMatrix[i,j]=JSD(as.vector(inMatrix[,i]),
+                              as.vector(inMatrix[,j]))
+     }
+   }
+   colnames -> colnames(resultsMatrix) -> rownames(resultsMatrix)
+   as.dist(resultsMatrix)->resultsMatrix
+   attr(resultsMatrix, "method") <- "dist"
+   return(resultsMatrix)
+ }
> data.dist=dist.JSD(data)

I found this code in http://enterotype.embl.de/enterotypes.html

When the code below is inserted R gives me the coordinates to create a scatter plot without the coordinate to create the loading plot.

To be clearer:

I have same samples characterized by the % of different bacteria (the variables) present in them. I would like in the plot the samples and in the same time the bacteria in order to determine which samples are near to specific bacteria.

PCO.1<- pco(data.dist, k=2)
PCO.1

I tried the code plot(PCO.1) but I obtained the same results.. only the visualization of the samples.

Have a nice day!

Giffredo

[komal.rathi]

Umm, so you want a plot like this?

[Giffredo]

mm... no actually... I would like a plot without cluster (don't care about it mow) but I want in the plot the samples, like here labeled with black point, and also the bacteria/enterotype that determine the position in the plot of the samples... In this way I could say if a sample is colse to a specific bacteria or not..

[komal.rathi]

The plot you are talking about is essentially a clustering plot. It is same as the plot that I linked except for the circle.

Just a suggestion, you are posting answers instead of comments. When you want to reply to my comments, you can click on the add reply button below my comment and reply.

[komal.rathi]

By the way, the R package ade4 can do exactly what you want. You may have to play with it though. Some good tutorials are here.