Sub-typing patients by a single gene's expression in TCGA dataset.

Question

[morovatunc]

Hi,

We have experimentally found that given gene X causes some phenotype. To start exploring the underlying mechanism, we decided to use TCGA data as its freely available. So within the cancer patients can we simple divide two samples based on gene X expression.

The first thing that came to my mind was Z normalise X gene and get the [-inf, +1.96] and [196,+inf] patients and do differentially expressed gene analysis.

Pleas keep in mind that this is only for preliminary data. So we aim just to hand wave and explore. Any idea would make my day? Please.

Best regards,

Tunc.

[Sean Davis]

What do you want to know about the TCGA dataset? Are you interested in differences in patient groups, or are you interested in finding genes that behave similarly to your gene X?

Answer 1

[Michele Busby]

Do I understand correctly that you are dividing the samples by the expression of the gene (rather than observing the expression of the gene in different groups)?

If so, the first thing I would do is make a histogram of the log TPM and see if it is bimodal. Then if it is you will know where the cutoff is. If it's not bimodal it may be just a noisy gene. I would do it in GTEX for the the normal tissue as well.

[morovatunc]

Yes. You are exactly right. Those samples are all spesified as Tumor. But I am trying to separate them based on the expression of a given gene (say X).

I havent thought about exploring the distribution and the GTEX thing. Thank you very much for reminding that ! I will give this method a try.

Answer 2

[Arup Ghosh]

If you are looking for gene expression patterns in Control/Tumor in different types of cancer, you can use FireBrowse.