RE: about data_mrna_seq_v2_rsem

[Ren, Danqing]

Hello Sir or Madam,

May I ask a question:
In the file titled data_mrna_seq_v2_rsem included into TCGA dataset, what is the data type of a certain gene mrna expression: FPKM, FPKM UQ or TPM?

How can I identify or clarify this information?
Thanks,

Dannie

 

 

From: Ren, Danqing
Sent: Tuesday, September 9, 2025 12:05 AM
To: cbiop...@googlegroups.com
Subject: about data_mrna_seq_v2_rsem

 

Hello Sir or Madam,

May I ask a question:
In the file titled data_mrna_seq_v2_rsem, what is the data type of a certain gene mrna expression: FPKM, FPKM UQ or TPM?

How can I identify or clarify this information?
Thanks,

Dannie

[Guizela Huelsz Prince]

Hi Dannie,

The data in data_mrna_seq_v2_rsem correspond to normalized counts (rsem.genes.normalized_results) from RSEM. You can find more details here: https://docs.cbioportal.org/user-guide/faq/#how-is-tcga-rnaseqv2-processed-what-units-are-used

Best,

Guizela

[Ren, Danqing]

Hello Guizela,

Thank you very much for your clarification. I just wanted to confirm one more point. From my understanding, the data_mrna_seq_v2_rsem file in the TCGA dataset corresponds to RSEM‐normalized results, which represent FPKM (Fragments Per Kilobase of transcript per Million mapped reads) values, rather than TPM or FPKM-UQ. Therefore, it would be appropriate to report the mRNA expression values in FPKM units when referring to specific genes.

 

Thanks again for your helpful response.

Best regards,
Dannie

 

From: Guizela Huelsz Prince <gui...@se4.bio>
Sent: Monday, September 22, 2025 5:03 AM
To: Guizela Huelsz Prince <gui...@se4.bio>; Ren, Danqing <Danqi...@uth.tmc.edu>; cbiop...@googlegroups.com
Subject: Re: about data_mrna_seq_v2_rsem

 

External: Increase caution when handling links and attachments.

[Guizela Huelsz Prince]

Hi Dannie,

The values in data_mrna_seq_v2_rsem are neither FPKM, nor FPKM-UQ, nor TPM. They are RSEM-normalized counts (rsem.genes.normalized_results), which are counts scaled using the upper-quartile percentile normalization (divide raw counts by the 75th percentile of nonzero counts for the sample, ×1000 as described here).

Best,

Guizela