CpGs with either fewer than 30 mapped reads or read counts >99th percentile (to exclude sites with potential PCR clonal amplification bias) were removed. CpG read depths across the 32 samples were normalized using the ‘normalizeCoverage' function in methylKit (Alkalin et al, 2012b). CpGs not represented in all samples and those mapping to sex chromosomes were not considered, leaving a total of 776 220 CpGs genome wide. Using this filtered and normalized set of CpGs, principal component analysis (PCA) was applied to identify potential sample outliers. Following outlier removal, we discarded CpGs at which the range in observed CpG methylation values (ie, maximum observed methylation value − minimum observed methylation value) within either group was ≥30% (n=208 914). At the remaining autosomal CpGs (n=567 306), logistic regression was used to test for differential methylation between THC and VEH groups. Observed P-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method (Benjamini and Hochberg, 1995).
