Mutual Exclusivity can anyone help clarify?

200 views
Skip to first unread message

wcharle...@gmail.com

unread,
Aug 23, 2018, 2:35:26 PM8/23/18
to cBioPortal for Cancer Genomics Discussion Group
I found some information in this group regarding the Mutual Exclusivity tab, but didn't find it super detailed. 

1: What scale is the log odds ratio? Log 10? And does this mean higher Log odds ratio means that there is a higher co-occurence of Gene A and Gene B in terms of alteration?
So if it were Log10 scale, this would mean that when I see Log Odds Ratio of 2.03 I would interpret how? that Gene A and Gene B have a 103-fold higher chance of being co-altered ? Not sure how to interpret this and put it into words. 

2: The object I'm really trying to acheive is look at correlation between 2 genes across cancer types ; is this possible? I want to say "In breast cancer samples, what are the levels of Gene B compared to Gene A?" with the hopeful goal of determining "Gene A and Gene B are negatively correlated. When Gene A is lower, Gene B is higher"  

Can anyone help?

Thanks!

Ethan Cerami

unread,
Aug 23, 2018, 3:04:44 PM8/23/18
to cBioPortal for Cancer Genomics Discussion Group, wcharle...@gmail.com
Hi There:

Thanks for reaching out.

When using the Mutual Exclusivity tab, I usually think of it this way…

Imagine you have two genes A and B.

If you have a positive odd ratio, that indicates that samples with
mutation in A also tend to have a mutation in B —> Co-occurence.
If you have a negative odds ratio, that indicates that samples with
mutation A almost never have a mutation in B —> Mutual exclusive

I tend to just look at whether it’s positive or negative, and then
jump to the p-value.  The p-value tells the likelihood that you would
see this pattern purely by random chance.  So, if you have a negative
odds ratio, and a p-value < 0.05, that means that you are observing a
mutually exclusive pattern, and that pattern is unlikely to occur
purely by chance, and therefore could be of biological or clinical
relevance.

The second part of your question has more to do with correlation.  For
example, if you are looking at the gene expression of Genes A and B,
you can enter both in the query both, and can click the “Plots” tab.
In there, you can select gene A for x axis and gene B for y axis, and
you could compare their expression and see if they are correlated.

I hope that helps!

Ethan

--
Ethan Cerami

On August 23, 2018 at 2:35:26 PM, wcharle...@gmail.com
> --
> 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 post to this group, send email to cbiop...@googlegroups.com.
> Visit this group at https://groups.google.com/group/cbioportal.
> To view this discussion on the web visit https://groups.google.com/d/msgid/cbioportal/aabc9c0f-9cac-4fe8-bb22-a22aaa409381%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>
Reply all
Reply to author
Forward
0 new messages