Obtain The Gene Level Cnv Matrix From Tcga Level3 Cnv Data

Question

[J.F.Jiang]

Hi all,

As TCGA offers great resources for multidimensional data, I want to obtain the gene level CNV matrix for some calculation. However, when I mapped the CNV region back to the refGene regions, I found too much genes that were covered by CNV regions, because some CNV regions were larger than 1M distance. The result will show up to 4K genes that have CNV, which is largely different from cBio portal.

So does anyone know better solutions? I just want to build a geneXsample matrix for CNV.

Thanks.

[akij]

Hi! how did you get the refGene regions from CNV regions. I also need gene level CNV matrix. How did you get that?

Answer 1

[Chris Miller]

In some samples (depending on the tumor type), I'd be surprised if there weren't several thousand genes in regions of copy number alteration. Whole-chromosome or chromosome arm amplifications and deletions are common in some cancers, and given that, your result sounds reasonable to me. Why, exactly do you expect fewer?

You'll also have to be more specific about what copy number data you're looking at on the cbio portal. Are they using recurrently-altered CN regions (across a cohort), as output by something like GISTIC?

[J.F.Jiang]

Thanks Chris. The genes covered by CNV in cBio is indeed calculated by GISTIC, which means the CNV regions must past some significance level. However, the input of GISTIC requires the level 2 data from TCGA. For those cancers using the SNP array, I can not have the permission to download unless I write a proposal to dbGap, which is now impossible for me.

So any idea to use the level3 data to obtain results like the GISTIC output?

[Chris Miller]

The core algorithm isn't that difficult to understand - it's essentially just stacking up the copy-number altered probes and asking which are altered recurrently. The stats for calculating recurrence get a little hairy to implement on your own though. I know you say getting access to level 2 data is impossible, but they're quite permissive as long as it's for research purposes - just need a short description of what you plan to do with it

[J.F.Jiang]

Thanks. I know if we can do that if we submit a proposal. However, get 3-4 cancers will not easy for me to get through.

Hope we can reach it.

[Na Sed]

@J.F.Jianig Could you figure out how to build [gene x sample] matrix for CNV. I have this problem, right now. Can you help me?