TCGA Isoform Expression Data

Question

[science03]

Hello everyone,

Is there anyone who could shortly enlighten me on how TCGA generates their isoform expression data briefly ? You should feel free to leave any informative links as well in the comments section. Thank you i n advance!

[i.sudbery]

Which collection of TCGA data are you interested in? The data on https://portal.gdc.cancer.gov/? The data on the Broad institute's firehose?

Some other source?

[science03]

I am interested in the isoform expression data on Firehose.

[GenoMax]

science03 : Please do not delete threads once they have received a comment/answer. If your question was solved then you should accept the answer provided below (green check mark).

Answer 1

[i.sudbery]

This is not easy information to find! The best explaination I've found is here: https://github.com/zyxue/MapspliceRSEM-clean-doc Which should be read in conjunction with this: https://bsbludwig.com/post/94066296740/what-do-tcgas-rnaseq-files-actually-show#fn:1

In short, the Broad firehose data is calcualted by RSEM (https://github.com/deweylab/RSEM). You get two columns per sample, raw_counts and scaled_estimate. THese are not the same names as actaully output by RSEM. raw_counts apparently corresponds to estimated_counts in the RSEM output and scaled_estimate is related to TPM: TPM is this value multiplied by 1e6.

The normalized files are these values multiplied by the 75 centile and divided by 300 (no idea why 300).

[igor]

I actually raised a related issue previously: What are RSEM normalized values?

It's reassuring to see that this is indeed not as simple to decipher as it may seem.

[science03]

Thank you so much for your help. I found the links pretty useful !