Performing eQTL on FPKMs?

[SamCurt]

We are planning to perform an eQTL analysis where the genotype input, at this time, will only be from a previous genotyping microarray run.

In this case, would it still be appropriate for us to run the analysis using a traditional QTL algorithm (R/qtl method em) and using Cufflinks FPKM as phenotype input? I see many papers talking about doing analysis this way using microarray data, but not so much using RNA-seq data, and are not completely sure if the results are directly comparable...

Thanks for the help!

[Simon Anders]

Well, at the very minimum, you should log-transform your FPKM values. But why don't you have a look at some RNA-Seq eQTL papers to see how they do it?

[SamCurt]

Well, at the very minimum, you should log-transform your FPKM values. But why don't you have a look at some RNA-Seq eQTL papers to see how they do it?

Hi Simon,

Not that I haven't did literature search before, but the answer is inconclusive. While I can generally conclude microarray methods can be applied to FPKMs (or in some cases raw counts), but the transformations used in different papers are different-- some papers would just do a simple log transform as you said, but some (like Pritchard) would do a qqnorm, then a PCA to remove 16 unknown factors, then qqnorm again...

I'm running a log transform to see what happens. Using raw Cufflinks FPKM gave unsatisfactory results.

Cheers,

John