FPKM vs TPM

23 views

Skip to first unread message

jtt1955

unread,

Feb 10, 2019, 8:52:49 AM2/10/19

to trinityrnaseq-users

Hi,

I just received comments back from a reviewer regarding FPKM vs TPM. I performed a fully replicated DE analyses using the trinity pipeline (RSEM, TMM normalized, edgeR) a few years ago which used FPKM values. More recently the pipeline now uses TPM. I have now been asked to redo my analysis to utilise TPM value as opposed to FPKM. Is this difference between FPKM and TPM regarding identifying differentially expressed transcripts significantly different?

Best, JTT

Brian Haas

unread,

Feb 10, 2019, 9:08:25 AM2/10/19

to jtt1955, trinityrnaseq-users

makes no difference. DE analysis is done using the counts, and not the transformed values. Also, note that TPM and FPKM values have a perfectly linear relationship, and after you do TMM (or other) cross-sample normalization (another linear rescaling), it shouldn't matter whether you used TPM or FPKM for making your heatmaps (afaik). People just like TPM because, before doing cross-sample normalization, the column values add up to 1M. After cross-sample normalization, that no longer is the case. ;-)

--
You received this message because you are subscribed to the Google Groups "trinityrnaseq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to trinityrnaseq-u...@googlegroups.com.
To post to this group, send email to trinityrn...@googlegroups.com.
Visit this group at https://groups.google.com/group/trinityrnaseq-users.
For more options, visit https://groups.google.com/d/optout.