See also
Comments on Wescott & Schloss 2017
UPARSE
The introduction and comparative validation overlooks several methods for making OTUs. In particular, W&S consider swarm but do not mention UPARSE (Edgar, 2013), which was shown to produce very accurate OTUs on mock community tests. As of 6/8/2017, UPARSE had 1,361 citations (per google scholar) while swarm only had 124. Also, UPARSE creates 97% OTUs, unlike swarm which constructs OTUs with varying identities. UPARSE is thus more highly cited than swarm and more directly comparable with OptiClust, so UPARSE should have been included.
Denoising algorithms
W&S also fail to mention denoising (error-correction) algorithms. While denoising per se does not explicitly create 97% OTUs, it is self-evident that it can be used as a pre-processing step for 97% clustering. In fact, an accurate denoising algorithm should be the preferred pre-processing step for 97% clustering because otherwise the input data for the clustering algorithm will have noise due to PCR and sequencing in addition to biological variation. Noise and biological variation are fundamentally different issues that are handled by different algorithms in state-of-the-art pipelines, but are not adequately distinguished by Westcott and Schloss.
DADA2 (Callahan et al. 2016) and UNOISE2 (Edgar, 2016) have been shown to be highly accurate in reconstructing correct biological sequences from noisy reads. These denoisers should also have been discussed and included in the validation.