Expression values for normal tissues

Question

[rizoic]

We have RNA seq transcripts per million(TPM) values for all genes in a tumor patient sample. However there was no sequencing done for normal samples from the same patient. We want to determine which genes are being over/under expressed in the tumor and having just its expression values is obviously not much informative.

Is there a resource or method that I can use to reasonably compare the tumor expression values with normal ones from a public resource for the same tissue?

I have tried to get median TPM values for the tissue from GTEx however though the TPM values are mildly correlated it seems unreliable to make any statement on the basis of this. I looked for the values for some house keeping genes mentioned in this publication but they are fairly different between the tumor and GTEx median.

Answer 1

[Kevin Blighe]

You can either go 'all out' and try to model highly / lowly expressed based on the level of expression of housekeeping genes (unreliable, as there is no good housekeeping gene), or you can simple transform your data to standardised scores (Z-scores) and take > 1.96 as highly expressed, and < -1.96 as lowly expressed. +/- Z=1.96 is equivalent to p=0.05.

Kevin

[rizoic]

Thanks. Z scores should give me a good idea about high/low expressed genes in the tumor. By chance are you aware of any resource to get an idea about the status of these genes in normal tissues?

[Kevin Blighe]

Apart from GTEx and FANTOM5, I am not sure. I believe there is some other large study ongoing (larger than GTEx), but its name eludes me at present...

The Z-score idea is also used by cBioPortal to gauge expressed / not expressed, by the way.