mutual exclusivity/co-occorrence analysis

[Lu Zhang]

Dear Sir or Madam,

I have a statistical question about mutual exclusivity/co-occurrence analysis of mutations in colon cancer on your website. Below is a link to the analysis done by the website for a subset of genes, I don't quite understand why some genes pair (e.g. HUWE1 and PRKDC) are labeled as tendency rather than strong tendency towords co-occurence with p vallue <0.05 and some gene pair (such as TP53 and WRN) was labeled as tendency towards mutual exclusivity with small p value and some (e.g. XRCC1 and RFC4) was labeled as strong tendency towards mutual exclusivity with big p value?

http://www.cbioportal.org/public-portal/index.do?tab_index=tab_visualize&cancer_study_id=coadread_tcga&genetic_profile_ids_PROFILE_MUTATION_EXTENDED=coadread_tcga_mutations&genetic_profile_ids_PROFILE_COPY_NUMBER_ALTERATION=coadread_tcga_gistic&Z_SCORE_THRESHOLD=2.0&case_set_id=coadread_tcga_3way_complete&case_ids=&gene_list=ercc6%0D%0Alig1%0D%0Arpa3%0D%0Arfc4%0D%0Atp53%0D%0Atp53bp1%0D%0Awrn%0D%0Aprkdc%0D%0Axrcc1%0D%0Ahuwe1&gene_set_choice=user-defined-list&COMPUTE_LOG_ODDS_RATIO=true&Action=Submit

 

Thanks,

Lu

---

UT Southwestern Medical Center
The future of medicine, today.

[Ethan Cerami]

Lu,

Thanks for contacting us.  I think one example should help answer your question.  If you look at RFC4 and XRCC1, you will see:


The portal does two statistical tests.  The first test is an odds ratio test, which simply tells you whether the events are mutually exclusive or co-occurrent, but does not tell you anything about statistical significance.

The second test is a Fisher's Exact test.  This tells you whether the observation is statistically significant.

In the example above, there two genes are strongly mutual exclusive (by the odds ratio), but, because there are not enough events, this could occur purely by chance, and the Fisher's Exact test therefore results in a large p-value.

I hope that explanation helps.

Ethan