Different results from CNV analysis resources both by Broad Institute?

Question

[nijironohata]

Question on CNV analysis using TCGA dataset.

In the page of firehose GDAC (http://firebrowse.org/api-docs/#!/Analyses/Amplified),

When I input “GNAS” in colon cancer, they respond it is not significant.

On the other hand, (http://www.broadinstitute.org/tcga/home)

When I input “GNAS” in this portal, they show the significant q-value in colon cancer.

Does anyone have an idea to explain why this difference happens?

Or does anyone have recommendation which pipeline should be used?

Answer 1

[agicict]

Though I did not analyze GDAC, judging from my experience, there are two possible reasons.

This may result from different calculation approach for p.value. For example, there are two popular statistical methods to compare two samples; one is t.test and the other is wilcoxn.

So you need to check whether they use different methods.

Additionally, this may be caused by different cohorts. You need to check if they were exactly same cohort (i.e same patients and information).

[nijironohata]

Thank you for helpful comment!