Aims. Polycystic ovarian syndrome (PCOS) is a common endocrine disorder characterised by ovarian morphological, systemic biochemical, and menstrual changes. Women with PCOS are at significantly increased risk of raised fasting glucose, impaired glucose tolerance, and diabetes. Recognition of these complications and early intervention are key to good health outcomes. We sought to identify DNA methylation patterns that may predict future diabetes onset in this high-risk PCOS population. Patients and Methods. Peripheral blood samples from women with PCOS and women with PCOS who later developed diabetes, were analysed by Illumina HumanMethylation450 BeadChip-arrays. Bisulphite-Pyrosequencing™ was used to validate and confirm array methylation data. Results. Array analyses identified 273 differentially methylated CpG loci (≥0.2 β-value change) at initial diagnosis of PCOS, between women who did or did not later develop diabetes. 19 of these sites demonstrated differential methylation in the same direction in ≥five sample pairs. Methylation in three of the candidates (cg11897887, cg02819655, and cg25542007) showed the best concordance with corresponding array β-values, and, most clearly differentiated ‘cases’ from ‘controls’. Conclusions. We have identified novel methylation biomarkers that could predict future onset of diabetes in this high-risk population. Use of methylation analyses to identify women who are likely to develop diabetes at diagnosis of PCOS may facilitate timely lifestyle interventions to reduce future morbidity.