Table 4. Denoising of mtDNA COI amplicons generated from community samples of Collembola from two forest sites on the island of Tenerife with the pipeline of Yu et al. (2012).

	Site 1			Site 2
	MID7	MID8	MID9	MID10	MID11	MID12
Total read count
Step 1 (quality control)	10,413	10,581	8,416	7,043	10,635	10,737
Step 2 (PyNAST, 60%)	10,394	10,572	8,392	7,040	10,603	10,712
Unique read count
Step 2 (PyNAST, 60%)	7,040	7,306	5,822	4,752	7,492	7,443
Step 2 (USEARCH)	2,021 [12]	2,032 [13]	1,621 [16]	1,452 [5]	2,255 [15]	2,147 [30]
Step 3 (MACSE)	709	719	305	152	940	817
Step 4 (DNACLUST, 99%)	580	610	267	139	805	710
Step 5 (CROP, 98%)	69	52	52	65	103	98

Steps 1–3 represent reduction of unique sequence volume by denoising, while steps 4 and 5 further reduce unique sequence volume by the creation of summary clusters of sequences. See Yu et al. (2012) for a detailed explanation of each of the steps. Bracketed values in step 2 represent sequences inferred to be chimeras with the de novo chimera detection function UCHIME in USEARCH, all of which were removed in step 4 of PyroClean.