cBioPortal data with negative log2(FPKM + 1) values

Question

[LauferVA]

Hello, I am looking at data downloaded from cBioPortal directly.

The metadata for the mrna-seq files asserts the data are log2(FPKM + 1), but there values as small as -0.3 in the matrix.

I am thinking maybe they did log2(FPKM) + 1 instead of log2(FPKM + 1), or maybe they forgot the 1 altogether, who knows.

Has anyone dealt with this before, i.e. has it been reported for other datasets downloadable from cBioPortalData?

Thank you
VAL

[Kevin Blighe]

I would contact them. MSKCC is a reputable organisation but not immune to making grave errors.

[LauferVA]

Copy. Thank you, Kevin Blighe

Answer 1

[Zhenyu Zhang]

I don' know if they have updated their algorithm recently. A few years ago when I looked at cBio RNA-Seq data, I believe that was z-score normalization.

[LauferVA]

they still use this, but it is in a separate table. the rna_seq_v2_mrna column is described as log2( FPKM + 1 ) in the metadata

[Zhenyu Zhang]

Interesting. Anything called "rna_seq_v2" in the TCGA world usually means the raw input is legacy hg19 RSEM calls.

[LauferVA]

ya i think it is a mistake.

this is not the only dataset like this