DNAm from whole blood was quantified at 485,512 CpG sites using the Illumina Human Methylation 450k BeadChips at the Edinburgh Clinical Research Facility. Full details of the quality control steps have been described previously (Shah et al., 2014 (link); Zhang et al., 2018 (link)). Briefly, raw intensity data were background-corrected and normalised using internal controls. Samples with inadequate bisulphite conversion, hybridisation, staining signal or nucleotide extension were removed upon manual inspection. Further, probes with a low detection rate (p>0.01 in >5% of samples), samples with a low call rate (<450,000 probes detected at p<0.01), samples exhibiting a poor match between genotype and SNP control probes, and samples with a mismatch between methylation-predicted and recorded sex were additionally excluded. This left a total of 450,276 autosomal probes. In analyses comparing blood and brain DNAm signatures, the last blood measurement before death was used and models were adjusted for the interval between the blood draw and death (see Supplementary Table 1; mean interval: 2.5 years, SD: 1.5).