Alpha Diversity Metrics


Quick question about the metric choices for alpha diversity. Is observed_otus equivalent to observed_asvs? This would be clearer if the new nomenclature was used. In addition, it would be very helpful to post a guide of the alpha diversity metric choices and a short explanation of how they are calculated for those who are unfamiliar with all of the different measures.

Thank you!



Hi @vogelm,
Thanks for posting!

Yes. If your features consist of ASVs generated by dada2 or deblur, then yes.

This name is not observed_asvs for a couple reasons: first and foremost, because it is the name of the metric in skbio, which is the package that actually performs these alpha diversity calculations. Hence we need to keep the name to maintain consistency with the packages that QIIME2 wraps.

Second, because alpha does not really know what types of features are in your feature table. If these are ASVs from dada2 or deblur, then yes observed_asvs would be a nice name. But some users' features may be OTUs if they use vsearch for OTU picking. Other users' features might even be something else, like genes or metabolites if they are importing metagenome or metabolome data into QIIME2 for analysis. So observed_otus is just a generic term. observed_features would really be the most appropriate "catch-all" term to use a the "new nomenclature", but that's where reason #1 rears its head.

Besides, sequence variants are essentially 100% OTUs (if you do not consider denoising to be an essential component of what it means to be a sequence variant), so this term is technically not incorrect for most users of QIIME2 (yourself included, if you have ASVs).

See here. As alpha diversity calculations are being performed by skbio, the docs for skbio are the best source for this information. I have created an issue here to add this url to the help documentation.

I hope that helps!


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.