Unconfigured Ad

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • seeker
    Member
    • Jan 2011
    • 26

    #1

    edgeR MDS plot: any way to get genes contributing to axes out?

    Hi All,

    I'm working with edgeR and have a MDS plot that is interesting but want to know which genes contribute to the two axes. Any suggestions of how to get this out?

    Cheers!
  • MBWatson
    Junior Member
    • Jul 2014
    • 5

    #2
    Hello! I am also interested in determining which genes contribute to the axes in my MDS plot. Did you ever find a way to do this? Thank you!

    Comment

    • seeker
      Member
      • Jan 2011
      • 26

      #3
      No, I'm afraid I never figured it out. Perhaps this will spur others who might know how to share.

      Comment

      • dariober
        Senior Member
        • May 2010
        • 311

        #4
        Hi- I think you want to look at the rotations (or loadings) from a Principal Components Analysis (MDS is PCA where the data are a distance matrix).

        Let's see this example... You have 10 samples and 23 genes. Gene A and B are anticorrelated while C is highly expressed in samples 4, 5, 6, and 7. The remaining 20 genes do not show any pattern of expression:

        Code:
        n<- 10
        A<- 1:n
        B<- c(1, 1, 1, 10, 10, 10, 10, 2, 2, 2)
        C<- n:1
        k<- matrix(rnorm(n= 20 * n, mean= 5, sd= 0.5), ncol= n)
        dat<- as.matrix(rbind(A, B, C, k))
        rownames(dat)<- LETTERS[1:nrow(dat)]
        
        round(dat)
          [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
        A    1    2    3    4    5    6    7    8    9    10
        B    1    1    1   10   10   10   10    2    2     2
        C   10    9    8    7    6    5    4    3    2     1
        D    5    5    5    5    5    6    4    5    5     5
        E    5    5    5    5    5    6    6    5    4     5
        ...
        If you do a PCA and plot you see that the 10 samples form 3 groups, as expected:

        Code:
        pc<- prcomp(t(dat))
        plot(pc$x)
        
        # Or also:
        biplot(pc)
        For question "Which genes contribute to the separation", let's have a look at the rotations:

        Code:
        round(pc$rotation[1:10,1:2], 3)
             PC1    PC2
        A  0.408 -0.567
        B  0.797  0.581
        C -0.408  0.567
        D  0.068  0.031
        E  0.009 -0.018
        F -0.009 -0.006
        G  0.008 -0.042
        H -0.030 -0.014
        I -0.048  0.014
        J -0.009  0.015
        ...
        We can see that A, B, C contribute the most to the variation.

        (For this questions you might get better answers on the Bioconductor mailing list...)

        Hope this helps (and make sure I'm not missing anything...)

        Comment

        • seeker
          Member
          • Jan 2011
          • 26

          #5
          Thanks for that. I was also sent this link which explains why they do not provide this information.

          Comment

          Latest Articles

          Collapse

          • SEQadmin2
            Multiomic Liquid Biopsy: What DNA/RNA Panels Reveal That DNA Alone Can’t
            by SEQadmin2



            Tumors contain complex mixtures of cell populations that evolve in response to treatment and host-specific factors. Tissue biopsies provide only static snapshots of a tumor’s state. But what if we could follow that evolution over time using a minimally invasive technique?

            Liquid biopsies can detect tumor-derived material circulating in the bloodstream, providing a way to study treatment response, detect emerging resistance, and discover new biomarkers. Circulating
            ...
            10-14-2026, 09:55 AM
          • SEQadmin2
            New Genomics Technologies Take Aim at Long-Standing Limits
            by SEQadmin2


            Researchers using sequencing and genomics tools often have to make trade-offs. They can choose between speed or scale, short reads or long-range information, or targeted panels or a view of the whole transcriptome. New technologies that have been released this year are built to address those tough choices.

            We asked six companies the same four questions to learn about their latest products. The new technologies bring a lot to the table, including rethinking sequencing
            ...
            09-28-2026, 10:25 AM

          ad_right_rmr

          Collapse

          News

          Collapse

          Topics Statistics Last Post
          Started by SEQadmin2, Yesterday, 11:13 AM
          0 responses
          21 views
          0 reactions
          Last Post SEQadmin2  
          Started by SEQadmin2, Yesterday, 10:48 AM
          0 responses
          18 views
          0 reactions
          Last Post SEQadmin2  
          Started by SEQadmin2, 10-15-2026, 10:52 AM
          0 responses
          36 views
          0 reactions
          Last Post SEQadmin2  
          Started by SEQadmin2, 10-14-2026, 05:39 AM
          0 responses
          42 views
          0 reactions
          Last Post SEQadmin2  
          Working...