PCoA plot of all samples about emp

[wasade]

AFAIK we should be okay. The full QIIMEdb merges with new OTUs took ~25gb when I was monitoring, and should have been submitted to a 64GB queue regardless. Greg, do you have approx sample and obs counts there?

[gregcaporaso]

@wasade: approximate number of samples is 5-10k (still working out exactly what will be included). approximate number of otus is a little harder: we're currently about 2/3 through OTU picking on the first iteration of data (~4.5k samples) and have around 2 million OTUs, with singletons excluded. So, maybe 3m OTUs overall?

[cuttlefishh]

My how time as passed. As of Dec 2015, @wasade has figured out that whole biom-format thing and is working on faster UniFrac. @ElDeveloper and @mortonjt are working on faster Emperor.

Recent update from @wasade on faster UniFrac with smaller biom tables (optimization with large biom tables continues):

We can compute unweighted unifrac on 1000 samples 
(AG, random sample of 1k rarefied set, against gg 97%) 
in about 30s on a laptop. 50s if you account for parse of 
the input tree and table prior to bdiv. Peak total memory, 
including threads is <1GB.