Removing Noise From Pyrosequenced Amplicons

BMC Bioinformatics

Table 6 Accuracy of 3% OTU construction following application of the noise removal algorithms

Name	AmpliconNoise			SLP 1%			SLP 2%			DeNoiser
3% OTUs	Good	Missed	Noise	Good	Missed	Noise	Good	Missed	Noise	Good	Missed	Noise
Divergent	26	0	1	26	1	162	26	1	11	24	0	3
Artificial	34	0	1	38	0	109	32	2	10	36	0	6
Even1	57	0	131	60	2	174	39	19	65	59	1	21
Even2	57	0	89	59	2	137	42	15	53	58	3	22
Even3	58	0	125	65	0	197	43	17	75	58	3	23
Uneven1	58	0	101	60	0	144	47	10	59	54	7	24
Uneven2	48	0	35	51	0	74	38	10	26	46	6	6
Uneven3	49	1	46	51	2	97	41	10	45	48	6	12
Titanium
σ_s = 0.1	66	0	8	74	0	169	67	1	18	15	53	14
σ_s = 0.04	67	0	21

Reads classified as chimeric by comparison with the references were removed prior to complete linkage OTU construction at 3% sequence difference with the corresponding reference sequences. Each OTU was then classified as 'Good' if it comprised reference sequences and denoised pyrosequenced reads, 'Missed' OTUs are formed only from reference sequences, and those OTUs containing only pyrosequenced reads are designated 'Noise'. The numbers of classified OTUs observed for each data set and noise removal algorithm are shown below.

ISSN: 1471-2105