Table 3.

The accuracy of different computational methods on PacBio amplicon sequencing data from mock communities

Dataset	Method	Feature Type	Total	TP	FP	Singleton	Strains Detected
Zymo	DADA2	ASVs	29	29	0	0	8
	mothur	OTUs	24	8	5	11	8
	uparse	OTUs	8	8	0	0	8
HMP	DADA2	ASVs	51	51	0	0	17
	mothur	OTUs	29	16	4	9	16
	uparse	OTUs	16	16	0	0	16

The PacBio long-read amplicon sequencing data from the Zymo and HMP mock communities was processed by DADA2, mothur and uparse (as implemented in usearch). DADA2 identifies exact amplicon sequence variants (ASVs) with single-nucleotide resolution, while mothur and usearch identify OTUs, i.e. clusters of reads within a 97% similarity threshold. Total: The number of features (ASVs or OTUs) identified. TP: True positives. FP: False positives. Singleton: Features with just one read. In these datasets, the singleton OTUs output by mothur consisted of chimeras, sequencing errors and contaminants. Strains detected: The number of strains in the mock communities identified by each method.