Does upper-quartile normalization in Cuffdiff also normalizes by transcript length?

Question

[trakhtenberg]

Does upper-quartile normalization option in Cuffdiff (Cufflinks 2.2 version) also normalizes by transcript length?

Answer 1

[igor]

Cuffdiff uses FPKMs, so the values are already normalized by transcript length to begin with regardless of further normalization.

[trakhtenberg]

Thank you for the feedback. I guess part of my confusion is the sequence of operations itself. Theoretically, the upper-quartile normalization method should be instead, not after, standard FPKM normalization. Whereas Cufflinks manual states: “FPKMs and fragment counts are scaled via the ratio of the 75 quartile fragment counts to the average 75 quartile value across all libraries”, which sounds like you are saying that this is an additional rather than an alternative normalization option.

If so, in Cuffdiff this option utilizes both, the original normalization by total of millions of fragments as well as upper-quartile in the next step. Shouldn’t the upper-quartile normalization method, as described in the literature, use upper-quartile instead of, rather than in addition to, normalization by total of millions of fragments?

I posted what I think underlies this follow-up question under a different posting and with more details: Upper-quartile normalization option for Cufflinks and Cuffdiff