CSS percentile using QIIME.

[Beatriz Gil Pulido]

Hi, 

I normalized my OTU table before to perform the alpha and beta metrics. I used QIIME for the normalization and to generate the 2d_plots for PCoA (using weighted_unifrac). I tried to do it using R and a percentile 'p' parameter  is necessary to be defined. In R , 0.5 is used as a default but, what is the percentile used in QIIME? (i've read somewhere that they referred to that 'p' parameter as 'quantile'. Sorry if I've mixed up the terminology or concepts but I'm not familiar since I'm not statistician). 

Regards, 

Bea. 

[Jamie Morton]

Hi Beatriz,

I've contacted one of the developers of this method, so expect to hear a response shortly.

Best,

Jamie

[Sophie]

Hi Bea,

For CSS, p is calculated automatically by the cumNormStatFast metagenomeSeq function in R:

obj = load_biom(input_path)

 p = cumNormStatFast(obj)

 obj = cumNorm(obj, p = p)

write_biom(MRexperiment2biom(obj, norm=TRUE, log=TRUE), out_path)

You can view the p by using the -s option in normalize_table.py

Hope that helps!

Sophie