cufflinks --compatible-hits-norm

[deepsea]

Hi,

I am testing cufflinks with two different normalization methods: total-hits-norm and compatible-hits-norm. My understanding is that the only difference between these two methods is the FPKM denominator: one is the number of all hits, the other is the number of hits within the reference GTF. So the ratios of FPKM values calculated by two methods should be a single fixed number, which is the ratio of hits. But, the ratios I got is a distribution. Anybody have an idea of the exact meaning of compatible-hits-norm? Thanks!

[sdriscoll]

sounds fishy. i wanna try this out and see what results i get. I'll report back unless of course someone comes in an explains it.

[deepsea]

Oh I am very sorry! It is my problem: one set was processed slightly differently. :-(

Thanks, sdriscoll!

[sdriscoll]

indeed. i tried this today and the difference in FPKM values between the two normalizations was constant out to 4 decimal places.