Mapped reads were filtered as follows: any read with more than three mismatches was trimmed from the 3′ end to contain three mismatches, any read pair that contained a cytosine mapped to a reference sequence thymine was removed, and any read pairs that had more than three cytosines in the non-CG context within a single read was removed (possible non-conversion in bisulphite reaction). Read pairs were then collapsed to remove clonal reads potentially produced in the PCR amplification from the same template molecule, based on a common start position of read 1. The total uniquely mapped, non-clonal read number for each library, average coverage and total sequence yield are detailed in Supplementary Table 1.
