Selecting highest/lowest quartile of patients

152 views
Skip to first unread message

c.r.l.ou...@gmail.com

unread,
Apr 20, 2020, 8:59:46 AM4/20/20
to cBioPortal for Cancer Genomics Discussion Group
Dear cBioportal support team,

I am interested in making a query that selects the 25% highest expressing patients and 25% lowest expressing patients. Is there a way to do this?

I have managed to select and compare highest and lowest values with an expression cut-off and the EXP>2 EXP<-2 query, but now I am interested in selecting highest/lowest expressing patients in the way mentioned above. I hope this is possible!

Best regards,

Clara

Tali Mazor

unread,
Apr 20, 2020, 9:24:21 AM4/20/20
to c.r.l.ou...@gmail.com, cBioPortal for Cancer Genomics Discussion Group
Hi Clara,

We recently added a feature to study view that will help you to compare groups based on expression of a particular gene.

Go to study view for your study of interest (make sure it has expression data). Click on the "Charts" menu in the upper right of the page, and then select the "Gene Specific" tab. Here you can enter your gene of interest and select your profile of interest (mRNA expression). A chart will now add showing mRNA expression of your gene. Go to the chart (it will be at the bottom of the page), hover your mouse over the top right of that chart and select Compare Groups/Quartiles. This will bring you to Group Comparison with 4 groups, representing each quartile of expression. You can de-select the middle two groups, which will leave you with a comparison of the lowest and highest 25% of patients.

For more on group comparison, I encourage you to look at our tutorials (https://www.cbioportal.org/tutorials#group-comparison)

-Tali




On Mon, Apr 20, 2020 at 8:59 AM <c.r.l.ou...@gmail.com> wrote:

        External Email - Use Caution        

--
You received this message because you are subscribed to the Google Groups "cBioPortal for Cancer Genomics Discussion Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cbioportal+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cbioportal/732bf3d4-581d-4afb-a2d5-3e7069db8dd6%40googlegroups.com.

Tali Mazor

unread,
Apr 20, 2020, 5:32:25 PM4/20/20
to Clara Oudenaarden, cBioPortal for Cancer Genomics Discussion Group
Hi Clara,

If I understand correctly, you want to identify the set of patients that are in e.g. the top 25% of expression for multiple genes? You can do that, although it takes a few steps and you will need to be logged in.

First, repeat the process described above. When you get to the group comparison page, use the overlap tab to select your group of interest (e.g. samples with top 25% expression) and create a new group of those samples. In the current group comparison it will simply be a duplicate of the existing group, but when you create the group yourself it will save itself back to your user profile, and in the future you can access that group directly from study view. Repeat this process for all of your genes of interest.

Alternatively, if you already know what expression cutoff you want to use for each gene, you can apply a filter by selecting bars on the bar chart of expression in study view and create a group from there.

Either way, you can then use the "Groups" menu in study view to select the groups you saved, and enter group comparison with all of those groups. You can then use the overlap tab to create a new group which is made up of the overlap from each of your single gene groups.

You can see a more detailed description of creating groups from the overlap tab in the group comparison tutorial.

-Tali

PS - Please keep the google group cc'ed when you reply.


On Mon, Apr 20, 2020 at 9:36 AM Clara Oudenaarden <c.r.l.ou...@gmail.com> wrote:

Hi Tali,

 

Thank you for this clear and quick explanation for my problem. This is exactly what I was looking for, and almost serves what I need. Is it possible to do exactly the same but for a gene signature. For instance, combination of 5 genes, like you can also do in a normal query and you can have an overlap of patients?

 

Best regards,

 

Clara

Clara Oudenaarden

unread,
Apr 21, 2020, 8:24:36 AM4/21/20
to Tali Mazor, cBioPortal for Cancer Genomics Discussion Group

Hi Tali,

 

Thanks again. I tried this, but my created groups didn’t save, maybe I wasn’t logged in! I will try again. I am happy there is so much ‘hidden’ (or less obvious) functions in your website!

Reply all
Reply to author
Forward
0 new messages