Bisulfite treatment of DNA leaves methylated cytosines unaffected, while non-methylated cytosines are converted into uracils. Subsequent PCR amplification converts these uracils into thymines. For any given genomic locus, bisulfite treatment and subsequent PCR amplification give rise to four individual strands of DNA which can potentially all end up in a sequencing experiment (Supplementary Material). Mapping of bisulfite-treated sequences to a reference genome constitutes a significant computational challenge due to the combination of: (i) the reduced complexity of the DNA code; (ii) up to four DNA strands to be analysed; and (iii) the fact that each read can theoretically exist in all possible methylation states. Even though there are a number of excellent short read mapping tools available, e.g. Bowtie (Langmead et al., 2009), these do not perform bisulfite mapping themselves.
