CNV discrepancy PanCancerAtlas vs GDC

[LUCILE MARIE PAULE JEUSSET]

Hello,

I have noticed that the TCGA PanCancerAtlas datasets differ dramatically from the TCGA GDC datasets in terms of copy number variation (HOMDEL, HETLOSS, GAIN, AMP in the Oncoqueries). This is surprising given the exact same samples are analyzed, yet many of the samples that were previously reported to harbour shallow deletion of a given gene in the PanCancerAtlas now seem to exhibit normal copy number or gain or amplification of the same gene in the GDC version of the dataset. I have found this trend to be true of all the genes tested so far, in all the cancer types I have compared.

Could you please explain why the pipeline used to reanalyze the data in the GDC version so often produces a different result than the PanCancerAtlas pipeline? Is one pipeline more accurate than the other? Are their known biases in the process that may skew the results one way or the other?

Any insight into these discrepancies would be greatly appreciated.

Thank you very much,

- Lucile Jeusset

[Nikolaus Schultz]

Dear Lucile,

Thank you for reaching out.

I am not surprised that you see differences in copy-number between the older TCGA data (PanCancer Atlas and others) and data from the Genomic Data Commons. The data are derived from the same samples and raw data but processed using completely different methods (GISTIC for PanCancer Atlas, ASCAT for GDC). We do not have much experience with the ASCAT data and have never compared the two. Given the extensive literature based on the PanCancer and Firehose data (which is very similar), I would suggest you use those data for your analyses.

I hope this is helpful.

Niki.

-- 
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 visit https://groups.google.com/d/msgid/cbioportal/YQXPR01MB5217A2255C57ADB253AF8AFFF072A%40YQXPR01MB5217.CANPRD01.PROD.OUTLOOK.COM.