Results
Initial array data screening identified 273 CpG loci where ≥5 of the six matched PCOS ‘control’ and the PCOS with subsequent-onset diabetes ‘case’ sample pairs displayed differential methylation (β-value ≥0.2) (Figure 1. Hierarchical clustering and heatmap).
The most promising biomarkers were identified on the basis of differential methylation in the same direction (all either hyper- or hypo-methylated in PCOS control relative to PCOS with diabetes cases) in five or six (out of six) matched pairs. These criteria identified 19 putative methylation biomarkers (cg identifier list in Table S2 ). These CpG loci were validated using Pyrosequencing™ in all 12 array samples (Primer sequences in Table S3 ). Methylation in three of the candidates (cg11897887, cg02819655, and cg25542007) showed the best concordance with corresponding array β-values, and, upon visual inspection of plotted methylation data, most clearly differentiated cases from controls (Figure 2 ).