Header Leaderboard Ad

Collapse

PCA with cufflinks data

Collapse

Announcement

Collapse
No announcement yet.
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • PCA with cufflinks data

    I was wondering if folks had any advice on the simplest way to generate a principal component analysis or MDS plot with Cufflinks output. I could group all my Cufflinks output into a data matrix and then try to visualize it but I was wondering if there was a built in way in R (perhaps through a tool or package like cummerbun) Thanks -Rich

  • #2
    Originally posted by greener View Post
    I was wondering if folks had any advice on the simplest way to generate a principal component analysis or MDS plot with Cufflinks output. I could group all my Cufflinks output into a data matrix and then try to visualize it but I was wondering if there was a built in way in R (perhaps through a tool or package like cummerbun) Thanks -Rich
    Hi Rich,
    Awesome idea, and one that can be very easily implemented with cummeRbund. If you run a cuffdiff on all of your samples, cummeRbund will take care of aggregating the information into a SQLite db behind the scenes. The generation of the FPKM matrix is very simple once this is complete:

    Code:
    >library(cummeRbund)
    >cuff<-readCufflinks()
    
    >allGeneFPKMs<-fpkmMatrix(genes(cuff)) #You can of course use the transpose if you want samples instead of genes
    >genes.pca<-prcomp(allGeneFPKMs)
    >biplot(genes.pca) 
    
    
    >allIsoformFPKMs<-fpkmMatrix(isoforms(cuff))
    >isoforms.pca<-prcomp(allIsoformFPKMs)
    etc,etc,...

    Please let me know how this works out for you. It may be something that I would like to integrate into cummeRbund for the future.

    Cheers,
    Loyal

    Comment


    • #3
      A PCA or MDS plot function built in to cummeRbund would be an excellent feature!

      Comment


      • #4
        Can i also vote for a PCA component for cummeRbund?

        Comment


        • #5
          Hi all, both PCA and MDS are now built in to the development version of cummerbund. (PCAplot and MDSplot). Will become release version on October 2nd, but can be downloaded from BioC now...

          -Loyal

          Comment


          • #6
            perfect news Thank you.

            Comment


            • #7
              That's great, the plot generated by the MDSplot command is very useful. It would be nice to be able to color samples by arbitrary experimental conditions, for example, to be able to make all replicates of one sample the same color.

              It is easy enough to run each command in the MDSplot function manually and supply a different "names" vector to the geom color option:
              customnames<- c("Wildtype", "Wildtype", "Wildtype", "Mutant", "Mutant", "Mutant")
              p <- p + geom_point(aes(x=M1,y=M2,color=customnames)) + geom_text(aes(x=M1,y=M2,label=names,color=customnames)) + theme_bw()

              But it would be nice to have that capability built into the function.

              Comment

              Working...
              X