FIG 1 .

Comparison of OTU quality generated by multiple algorithms applied to four data sets. The nearest, average, and furthest neighbor clustering algorithms were used as implemented in mothur (v.1.37) (25). Abundance-based greedy clustering (AGC) and distance-based greedy clustering (DGC) were implemented using USEARCH (v.6.1) and VSEARCH (v.1.5.0) (3, 5, 26). Other de novo clustering algorithms included Swarm (v.2.1.1) (6, 7), OTUCLUST (v.0.1) (27), and Sumaclust (v.1.0.20). The MCC values for swarm were determined by selecting the distance threshold that generated the maximum MCC value for each data set. The USEARCH and SortMeRNA (v.2.0) closed-reference clusterings were performed using QIIME (v.1.9.1) (28, 29). Closed-reference clustering was also performed using VSEARCH (v.1.5.0) and NINJA-OPS (v.1.5.0) (16). The order of the sequences in each data set was randomized 30 times, and the intramethod range in MCC values was smaller than the plotting symbol. MCC values were calculated using mothur.