Additional file 1: Supplementary Methods. S1. QQ-plots for the results of the AD case-control EWAS (A, λ=1.0005), and the Braak stage EWAS (B, λ=1.0175) in the Oxford datasets. S2. Manhattan plot for the results of the AD case-control EWAS in the Oxford dataset; the red line indicates the experiment-wide significance threshold of 7.51E-08, whereas the purple line indicates the suggestive significance threshold of 1.00E-05. CpGs with suggestively significant association are marked in purple and annotated with the gene name according to the Illumina manifest (v1.0 B5). NB that three CpGs (on chr. 8 and chr. 14 were not annotated to any genes). S3. Manhattan plot for the results of the AD Braak stage EWAS in the Oxford dataset; the red line indicates the experiment-wide significance threshold of 7.51E-8, whereas the purple line indicates the suggestive significance threshold of 1.00E-05. CpGs with suggestively significant association are marked in purple and annotated with the gene name according to the Illumina manifest (v1.0 B5). NB that one CpG (on chr. 14 was not annotated to any genes). S4. QQ-plots for the results of the AD case-control meta-EWAS (A, λ=1.16), and the Braak stage meta-EWAS (B, λ =1.24) with fixed-effect models across three datasets (London-1, London-2, Oxford). S5. QQ-plots for the results of the AD case-control meta-EWAS (A, λ=0.92), and the Braak stage meta-EWAS (B, λ =1.00) with random-effect models across three datasets (London-1, London-2, Oxford). S6. Scree plots for the results of a principal component analysis (PCA) on DNAm using binned CpGs in the Oxford (A, n=149 samples), London-1 (B, n=104 samples) and London-2 (C, n=92 samples) datasets. For the main EWAS analyses reported in this paper we selected n=13, n=13, n=15 PCs for the Oxford, London-1, and London-2, datasets, respectively, to account for known and unknown confounders of DNAm in these samples.