Figure 1

Study design and analytical approach to identify longitudinal differential methylation preceding type 1 diabetes. (a) We measured the proportion of DNA methylation (Beta, ranging 0 to 1) on one to five pre-disease samples for 87 cases (orange) and 87 frequency-matched controls (blue) from DAISY. Selected samples included cord blood (birth, 1), infancy between 9 and 16 months (2), the last disease-free sample prior to IA onset (3), the first sample after IA onset, when IA was first detected (4), and the last sample prior to T1D diagnosis (5). (b) We examined longitudinal differences in the rate of methylation change between cases and controls and the average methylation difference between cases and controls. Linear mixed models estimated the longitudinal difference in changing methylation, and the average longitudinal difference in methylation (change in the meta-analysis beta—“delta meta beta”) between T1D case and control groups. Results from probes measured on both 450 K and EPIC platforms were combined using a meta-analysis. (c) To examine the impact of preclinical autoimmunity at candidate probes, we examined cross-sectional differences in methylation at three important periods in the natural history of T1D: cord blood (1), pre-IA onset (2 or 3), and post-IA onset (4 or 5). Cross-sectional lookup analyses included 37 cases and 35 controls at cord blood, 43 cases and 44 controls at pre-IA onset, and 85 cases and 85 controls at post-IA onset. (d) Analyses were performed on both the position and region level. (e) The majority of probes on the 450 K array were also present on the EPIC array.