search

rnaseq / degraTrans

RNA-seq performances in the context of degraded transcriptome annotations.
0 stars 0 forks source link

Expression measure #7

Open rnaseq opened 8 years ago

rnaseq commented 8 years ago

Select the expression measure(s) to use: e.g. FPKM, counts, TPM?

Add consideration re. is it easy to extract these from all of the tools?

rob-p commented 8 years ago

The only measure above to which I object is FPKM. If we want to measure relative abundance, we should use TPM (there's no good argument favoring FPKM over TPM). It may make sense, however, to consider both counts and TPM, as they convey somewhat different information (e.g. how well, or poorly, a method is correcting for feature length).

blahah commented 8 years ago

I agree - there is never a reason to use FPKM :).

TPM, raw (estimated) counts and perhaps effective counts?

mw55309 commented 8 years ago

Hi All

Happy to drop FPKM (though I bet you all £5/$5 a reviewer suggests we put it in ;-))

If we are using cufflinks, we need to decide (a) which tool to use to produce the expression measure (cufflinks, cuffquant, cuffnorm) and (b) whether to use (normalised) counts or some form of transformed FPKM.

Can we also explicitly state the formulae for:

TPM calculation from counts TPM conversion from FPKM (if needed)

rnaseq commented 8 years ago

I also agree re. dropping FPKM and using raw counts and TPM measure.

'TPM calculation from counts' and 'TPM conversion from FPKM' are stated in this nice blog post: https://haroldpimentel.wordpress.com/2014/05/08/what-the-fpkm-a-review-rna-seq-expression-units/

rob-p commented 8 years ago

@mw55309, I'll take the bet if I pay $ and you pay £ ;P. @Blahah, what is the difference here between "estimated" and "effective" counts? Are effective counts what you get here when you simply distribute "counts" to transcripts based on nucleotide fraction or NPM (nucleotides-per-million)?