See also
Comments on Wescott & Schloss 2017
I attempted to reproduce the procecedure described in the W&S17 paper and ran the OptiClust algorithm on reads of a mock community. Data and mothur scripts to reproduce my results can be downloaded here. I used mothur v1.39.5.
Reads are subsets from Bokulich et al. 2013 dataset #2, which I chose because it has no cross-talk (because no other samples were included in the run) and contaminants have very low abundance. Almost all reads can be accounted for by the designed strains in the community.
These are V4 reads of two Even and two Staggered HMP mock communities (22 strains). The stag1k subset has 1,000 reads of one staggered sample, the bok10k subset has 40,000 reads, 10k for each of the four samples.
On stag1k, OptiClust reported 276 OTUs from OptiClust, and on bok10k OptiClust reported 5,757 OTUs.