All sequence alignments were performed against the NCBI36/hg18 human reference. Single-read MethylC-Seq sequences were processed and aligned as described previously18, except an additional filter was added to remove any mapped reads in which a read-C base was aligned to a reference-T base. Paired-read MethylC-Seq data was mapped and processed as described previously18 with the following modifications to accommodate the paired-read data-type. Both reads in a pair were trimmed of any low-quality sequence at their 3′ ends and mapped to the reference genome with Bowtie v. 0.12.530 in paired-read mode, using the following parameters: -e 90 -l 20 -n 0 -k 10 -o 4 -I 0 -X 550 -pairtries 100 -nomaqround -solexa1.3-quals. Mapped reads in a read pair that overlapped were trimmed from their respective 3′ ends until the reads no longer overlapped, leaving a 1-bp gap.
