BS Seeker: precise mapping for bisulfite sequencing

BMC Bioinformatics

Table 1 Mapping 1M synthetic human chr. 21 reads onto human chr. 21

Aligner	Experimental protocol	Uniquely (or best^a) mapped reads^b(%)	Accuracy (%)	Methylation rates^c(CG/CHG/CHH) (%)			CPU time
No base calling error
BS Seeker	Lister et al	91.7	100	72.0	0	0	209 s ^d
BSMAP	Lister et al	92.1	100	72.3	0	0	15h43m20s
RMAP	Lister et al	91.7	100	72.0	0	0	185 s
MAQ	Lister et al	>99.9	93.4	67.7	0	0	353 s
BS Seeker	Cokus et al	89.6	100	72.0	0	0	263 s ^d
BSMAP	Cokus et al	89.8	99.6	72.4	0	0	15 h 46 m 40 s
RMAP	Cokus et al	80.2	99.0	71.3	0	0.1	400 s
MAQ	Cokus et al	73.0	92.2	68.2	0	0	665 s
Simulated base calling errors
BS Seeker	Lister et al	91.2	99.54	71.5	0.4	0.4	217 s
BSMAP	Lister et al	91.1	99.57	72.1	0.4	0.4	15 h 19 m 51 s
RMAP	Lister et al	91.0	99.52	71.6	0.4	0.4	188 s
MAQ	Lister et al	99.5	92.9	67.8	0.4	0.4	340 s

^a BS Seeker, BSMAP, and RMAP discard non-uniquely mapped reads. MAQ keeps non-uniquely mapped reads and assigns them to one of the best-matching positions. ^b Up to 2 mismatches are allowed. ^c The simulated methylation rates are set to be 72%, 0%, and 0% for CG, CHG, and CHH. ^d Preprocessing reference genome needs 2-5 cpu minutes for the first run.

ISSN: 1471-2105