Fig. 5

Analysis of age-associated DNAm patterns within individual BBA-seq reads. a Heat map to exemplarily depicts frequencies of DNAm patterns within the 36 neighboring CpGs of the ELOVL2 amplicon in BBA-seq data of a young (21 years old) and old sample (72 years old). b For comparison, heatmaps are presented based on random simulation of DNAm patterns under the assumption that DNAm at neighboring CpGs occurs entirely independent (simulations correspond to 21 and 72 year old donors). c Pearson’s correlation of DNAm levels between neighboring CpG sites within ELOVL2 amplicon (BBA-seq data of training set). d For each BBA-seq read of ELOVL2 training set, we estimated the epigenetic age based on the binary sequel of methylated and non-methylated CpGs. The plot depicts relative read count of every donor in the training set that were classified for predicted ages between 0 and 200 years (relative read count normalized by read count per sample). e–g The mean age-predictions based on individual BBA-seq reads were determined for each sample and then plotted against the chronological age of the samples of the training (blue, n = 38) and validation set (red, n = 39). This analysis was performed independently for the amplicons of ELOVL2 (e), PDE4C (f), and FHL2 (g)