calculations for z-scores not matching cBioportal

[Seb ...]

hi all

I downloaded the latest BRCA RSEMv2 RNASeq data and calculated z-scores with the formula:

z  = ((value tumor)-(mean value normal)) / standard_deviation(normal)

normal and tumor were defined by using the TCGA barcode and I downloaded my data from TCGA/Broad Firehose like yesterday.

however, the values I get aren't as close as those in cBioportal as I'd wish them to be. this is affecting my ability to detect with a z > +/- 2 samples that are altered vs. not altered...which in turn affect survival analyses.

I know cBioportal uses ploidity to define "normal" but I'm not sure where I can get those info from.

any suggestions on how to fix this?

thanks

[Nikolaus Schultz]

We do not currently use normal samples for the z-scores calculation, only tumor samples, and only the subset that is diploid for a given gene. The diploid fraction is defined by the GISTIC copy-number data (all_thresholded_by_gene.txt, 0 = diploid).

I hope this helps.

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.
To post to this group, send email to cbiop...@googlegroups.com.
Visit this group at https://groups.google.com/group/cbioportal.
For more options, visit https://groups.google.com/d/optout.