Why is single_rarefaction.py ok for many QIIME analyses?

[abir...@gmail.com]

Hi, All,

I'm planning to rarefy some data so that I can normalize for effects of different library sizes. I had originally assumed that I'd need to use multiple_rarefactions_even_depth.py to generate numerous different rarefactions at my chosen depth and then somehow combine them, but I see that (a) many diversity analyses in QIIME use a single rarefaction (see https://groups.google.com/forum/#!searchin/qiime-forum/rarefaction/qiime-forum/bgEp9dNrnN0/LC_kPPuNCAAJ ) and (b) there isn't existing functionality in QIIME or biom to combine multiple rarefied OTU tables at the same depth together again (see https://github.com/biocore/qiime/issues/287 ).

I'm worried about doing just a single rarefaction to my chosen depth because, well, what if that one rarefaction happens to grab a weird, unrepresentative subsample of the data?  Won't that throw off my whole subsequent analysis?  However, given the above info that a single rarefaction is considered enough for a lot of analyses and that there isn't even enough need to combine multiple rarefactions (at the same depth) to build out this functionality, I'm assuming the community knows something I don't about why my concern is unnecessary.  Can anyone explain why or point me to a reference to put my mind at ease? :)

Thank you,

Amanda

[Kyle Bittinger]

The script jackknifed_beta_diversity.py can run the beta diversity analysis many times for a consistent number of reads per sample.

http://qiime.org/scripts/jackknifed_beta_diversity.html