Z score for from RNA Seq V2 RSEM

[cail...@gmail.com]

Hi,

I have downloaded rsem_gene_normalized data from firehose and generated box plots comparing tumor vs normal myself. But the trend doesn't seem to be consistent with up/down-regulation of mRNA expression from cbioportal.

For example, for BRCA TCGA provisional data, from my own analysis KCTD14 mRNA is clearly down regulated in tumor though it is amplified by GISTIC result, the Z-score from cbioportal shows that the mRNA levels are up-regulated.

For datasets like BRCA for which normal samples are available, are z-scores calculated using normal tissue as a reference population, or all diploid tumors? Are z-score calculated by pooled samples or paired samples.

Another example that the z-score doesn't quite make sense is attached:

CDKN2A is a tumor suppressor gene. It is deleted and mutated but mRNA z-score shows it's upregulated in LUAD...

I wonder if the mRNA up/down regulation should be calculated with a better method...or do I have any problems interpreting the data.

Thank you

Ling

[Nikolaus Schultz]

Hi Ling,

The z score for a gene is calculated using tumor samples diploid for the gene as reference (all pooled samples and not pairwise comparisons) - we do not currently use normal samples. Therefore you may see different results when compared to what you would get out of a tumor / normal comparison.

CDKN2A is a tricky gene. Its expression is induced in RB-defective tumor samples, so it may appear to be up-regulated in a fraction of tumor samples. 

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.
For more options, visit https://groups.google.com/d/optout.
<CDKN2A_example.JPG>