DNAm Data Preprocessing
Raw signal intensities were imported from.idat files into the R statistical environment31 and converted into beta values (the proportion of DNA methylation at individual sites was measured) with the bigmelon package.32 These data were processed via a standard pipeline including the following steps: (1) detection of outlier samples via principal-component analysis and Mahalanobis distance equivalents, (2) confirmation of complete bisulphite conversion via control probes, (3) comparison of estimated age from the data via the Horvath Epigenetic Clock algorithm33 and reported age at sampling, and (4) visualization of principal components. Data were normalized with the dasen function within the wateRmelon package,34 which performs background adjustment and between-sample quantile normalization of methylated (M) and unmethylated (U) intensities separately for type I and type II probes. Samples that were dramatically altered as a result of normalization were excluded on the basis of the difference between the normalized and raw data; those with a root mean square and standard deviation > 0.05 were removed. Samples were then filtered so that those with >1% of sites with a detection p value > 0.05 were excluded. Finally, DNA-methylation sites with a bead count <3 were excluded along with those in which >1% of the sample had a detection p value > 0.05. The raw DNA methylation data from the final sample set was then re-normalzsed with the dasen function. The final dataset included 857,071 DNA-methylation sites and 1,175 individuals for subsequent analysis. These DNAm data are available upon request through the European Genome-Phenome Archive under accession code EGAS00001001232.