What are RSEM normalized values?

Question

[igor]

I keep seeing people refer to RSEM-normalized RNA-seq values. I am trying to understand what exactly those are. Based on the RSEM website, it sounds like RSEM is only calculating the raw counts and then using EBSeq for normalization. However, the original RSEM paper mentions a normalization strategy, so maybe RSEM is doing the normalization itself. Which is the correct version?

As a followup, if I want to calculate values comparable to what might be expected with RSEM, what would be my best option?

Answer 1

[Michael]

I think that refers to TPM (Transcripts per Million), see http://deweylab.biostat.wisc.edu/rsem/rsem-calculate-expression.html#output

[igor]

That's a reasonable assumption, but then they would just say TPM or FPKM (since RSEM outputs both). Fo example, when someone use Cufflinks, they just say FPKM values.

That link mentions *.genes.results files. However, RSEM also has *.genes.normalized_results (according to https://wiki.nci.nih.gov/display/TCGA/RNASeq+Version+2).

[Michael]

The people should be more specific, I agree when communicating these results. However, the values specificly computed only by RSEM are TPMs. Also people should say whether they refer to summarized counts with respect to transcripts or genes because rsem makes both.