Clinical Covariates And Level 1 Data In The Cancer Genome Atlas (Tcga)

Question

[Ryan D]

We are interested in using the data available through The Cancer Genome Atlas (TCGA) to look at pharmacogenomic outcomes. Essentially, we want to look at how patients with a type of cancer (e.g. breast cancer) responded, or failed to respond, to medication used. There are 3 levels of data available for SNP and CNV genotype data. The clinical treatment data is apparently sparse with a lot of missing variables or incorrectly annotated. I wonder if anyone has experience with this or could comment on the correctness or completeness of this dataset and the feasibility of examining pharmacogenomic outcomes.

I have had less luck getting a straight answer to this question from the source and the application process is a bit of a pain. Is there someplace this information can be found? Or do any of you know?

Answer 1

[Tcga]

Why not actually just write directly to them? [e-mail]

[Ryan D]

I have written to them. They wrote back telling me to use the data matrix to answer these questions. https://tcga-data.nci.nih.gov. After downloading it, I was astonished to see the amount of data missing. It looks like the kind of exploration we intended is not possible--or at least not easy--using TCGA.

On the plus side, their data matrix does appear quite useful. It would just be nice if they could relax their iron-clad grip on level 1/2 data to make it a little less annoying for researchers to get to.

Answer 2

[J.Rodrigo Flores]

Hi Ryan,

Me and some collegues worked with TCGA data about a year ago. Specifically with glioblastoma samples. Certainly the amount of missing data depends on the cohort you want to work with. The experience in our case as far as I remember was that although for some samples there's missing clinical data as you say, the majority do have a lot of details regarding treatments, type of treatments, periods of exposure, etc. I'm not quite sure but atually I think this is one of the best public data repositories you will find with such an extensive clinical annotation.

In my opinion the big downside of working with TCGA is understanding all the details of the way the deliver data. It may be a bit of a pain at first.

J. Rodrigo