Evaluating the accuracy of amplicon-based microbiome computational pipelines on simulated human gut microbial communities

Table 1 Species Level Classification

Pipeline	OTU Strategy	OTU algorithm	Reference	Undercalled (%)	Undercalled (Ranks off)	Correct (%)	Misscalled (%)	Miscalled (Ranks off)	Lost (%)
QIIME	Closed		GreenGenes	55.8	1 (1–4)	18.9	22.3	4 (1–10)	3.0
QIIME	Sub	UClust	GreenGenes	63.3	1 (1–3)	12	24.5	4 (1–10)	0.2
QIIME	Sub	UClust	Silva	77.1	1 (1–6)	8.8	13.8	4 (1–14)	0.2
QIIME	De novo	UClust	GreenGenes	61.4	1 (1–3)	12.2	26.2	4 (1–10)	0.1
QIIME	De novo	UClust	Silva	77.7	1 (1–3)	8.7	13.4	4 (1–12)	0.1
QIIME	De novo	Swarm	GreenGenes	61.5	1 (1–4)	12.4	25.9	4 (1–10)	0.1
MOTHUR	Closed		Silva/RDP	54.6	1 (1–3)	6.9	21.9	10 (4–12)	16.6
pplacer	De novo	Swarm	RDP	68.2	1 (1–8)	18.1	12.5	4 (1–10)	1.2

Summary of Classification Performance. On the left are the various conditions tested. The first column specifies the pipeline, the second the OTU strategy, the third the methodological details (e.g. reference set or algorithm used). Table 1 is for species-level classification, Table 2 is for genus-level. Source organisms can be correctly called, undercalled (in the correct clade, but not the target species or genus level classification), or miscalled (placed down the wrong taxonomic clade). We present both the percentage in each category (correct, undercalled, and miscalled) and the median (min and max parenthetical) taxonomic ranks off for underacalled and miscalled source organisms

ISSN: 1471-2105