Correlation analysis and clustering

41 views
Skip to first unread message

Maarten van Gils

unread,
Dec 3, 2013, 1:45:09 PM12/3/13
to gitool...@googlegroups.com
Dear all,

I am using the correlation tool in Gitools to correlate gene expression changes (SLRs) of several genes (n=130) among each other and it seems to work perfect. However, If I would like to cluster (eg the rows of) the correlation matrix (which worked fine in previous editions I think) I experience some problems. The data look nicely clustered, the order of the genes names on top of the figure (for the columns) changes, however, the order of the genes name on the right side of the figure (rows) does not change. This means (as far I am aware) that the correlation matrix does not longer provide reliable results. Based on the design of the figure, the order of the genes should be the same for both axes in order to obtain the correlation coëfficiënts of 1 at the edge of the matrix. I hope someone can help me with this issue?

Thank you very much.

Michael P Schroeder

unread,
Dec 4, 2013, 5:53:31 AM12/4/13
to gitool...@googlegroups.com, vangils...@gmail.com
Hi Maarten

First, thanks for taking the time share the issue. It seems that you have stumbled upon a bug: The item order resulting from the clustering is not applied to the other dimension (row/columns). Do you remember by any chance with which version it was working last? If you do you could try to open the correlation with the older version ( which you may find here: http://gitools.org/downloads/).

Alternatively, there is a workaround with the actual version by exporting the matrix into a flat text file and opening it as normal heatmap which will show the values for all cells (squared heatmap instead of triangular). If you want to try the workaround, follow the steps below:
- Select the menu File -> Export -> Export Matrix.
- A dialog will ask you which value you want to export: select Score.
- Save the file on your computer with the extension .cdm. For example mycorrelation.cdm
- Once saved, select the menu File -> Open -> from filesystem and then your mycorrelation.cdm
- You have now the same data in a heatmap, but each data point is shown twice in a squared heatmap instead of the triangle view in a correlation result.
- Apply your clustering to one dimension, f.ex. rows. 
- Select all rows, right click on the selection and choose the option "Copy selected labels to clipboard"
- Select the menu Filter -> by annotations
- Choose columns and paste the copied row ids into the text field (CTRL+V) and hit ok.

You should now have the correlation result ordered according to the clustering in both dimensions. Note that if you would like the same color scale as in the correlation results, maximize Settings on the left bottom and select the Correlation scale from the drop-down selection.

I hope I could help and we will address the bug that you discovered.

Best,
Michael



--
You received this message because you are subscribed to the Google Groups "Gitools users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gitools-user...@googlegroups.com.
Visit this group at http://groups.google.com/group/gitools-users.

Reply all
Reply to author
Forward
0 new messages