Quantile normalization for RNA seq data?

[Boel]

Dear All,

I would like to implement quantile normalization on my RNA seq data (as it has been shown to reduce bias in DE calls, Bullard et al. BMC Bioinformatics 2010, 11:94). But my data is already normalized since I use Cufflinks. Has anybody thought about or even implemented quantile normalization into a TopHat/Cufflinks-like pipeline?

Would it be completely statistically or mathematically incorrect to quartile normalize the FPKM values of all transcripts?

Any input would be appreciated,
Boel

[Cole Trapnell]

We're planning to do this in an upcoming release of Cufflinks

[Boel]

That's great. Might I ask how you are planning to do this? And how far away this release of Cufflinks is?

[Cole Trapnell]

We haven't worked out all the details yet, so I won't comment on the "how" parts yet, except to say that we're consulting with Elizabeth Purdom on how to adapt some of their contributions to Cufflinks. However, we believe it's mathematically straightforward to adopt quantile normalization to our statistical model - the bottleneck will be the engineering, because there will have to be some code reorganization, etc.

As far as when this would make it into the wild, I can't really say. There are a number of new features that I really want to get in place. I would estimate the time scale as between several weeks to several months, depending on what else comes up in the meantime.