Otu Percentage Identity Related To Taxon Level

Question

[Daniel]

Hi

I have been generating OTUs at multiple different identity levels (via qiime) to see the effect on otu numbers and diversity and I want to relate it to taxonomic levels.

In the past I have generally followed that roughly (ROUGHLY):

• 97% ~ Genus
• 94% ~ Family
• 88% ~ Order

But I wouldn't put my want to sign my name to that. Does anyone have any experience or references to this effect?

Thanks

Answer 1

[cts]

It's a tough question to answer since bacteria taxonomy tends to be subjective and there are no globally defined cutoffs set down for taxonomic delineation apart from the 'rule of thumb' that you mention. I can't think of any paper that goes though this conclusively but you could look at:

Konstantinidis and Tiedje (2005) "Towards a Genome-Based Taxonomy for Prokaryotes" J. Bacteriol. 187(18):6258–6264 doi:10.1128/JB.187.18.6258–6264.2005

Particularly figure 3 where they compare the 16S identity to the overall genome sequence identity, overlapped with the taxonomy of the genomes. Another paper to look at is an old one by Stakerbrant and Goebel which describes a couple of pitfalls in 16S analysis, namely that high 16S identity does not necessarily correlate with whole genome identity and that the diversity is not uniformly distributed. There was also a paper that came out just last month by Peer Bork's group that also has some interesting points about bacterial speciation.

So most of these papers only look at the species level (ie 97%), the higher order classifications don't seem to get talked about as much. These cutoffs though are based on the full length 16S sequence, so using those metrics on an OTU is dubious. This paper and others point out systematic biases in diversity metrics based on the primer set you use. It's all very depressing really.

Answer 2

[Rob]

Here's an analysis of 6,633 RecA proteins that shows these numbers are off somewhat.

https://edwards.sdsu.edu/research/percent-similarity-at-different-taxonomic-levels/