A preprint classified age-related DNA methylation patterns in blood into ten types
A preprint classified age-related DNA methylation patterns in blood into ten types
In a preprint posted on 29 August, the authors mapped age-related changes in methylation, a chemical modification of DNA. Using 3 501 whole-blood samples, they confidently assigned more than 600 thousand of 811 870 CpG sites, short regions of DNA where methylation is measured, to ten pattern types. The authors linked one type to an inflammatory signal and mortality.
The same age-related shift in methylation can take different forms. At some sites, the mean level changes gradually. At others, variation between individuals increases. Still others form groups associated with sex or DNA variants. To distinguish these patterns, the authors generated an image for each CpG site, with age on one axis and methylation level on the other. A neural network identified the shape of each pattern and assigned the sites to ten types.
Changes in the proportions of immune cells in whole blood alter the overall methylation level. An analysis of paired blood samples collected ten years apart found that changes in immune cell proportions explained most of the shift in a 48-site methylation measure. To separate this effect, the authors selected 399 CpGs for NoiseClock across three independent cohorts after statistically adjusting for the estimated proportions of twelve cell types. The mean methylation level at these CpGs did not change with age, but variation between individuals increased. The researchers then evaluated NoiseClock, defined as the mean deviation of these marks from their usual levels, in 14 blood datasets and five datasets of sorted immune cells. Its association with age persisted.
NoiseClock was also associated with a methylation-derived measure of inflammation after adjustment for cell composition. In the Scottish Generation Scotland cohort, it was calculated for 18 859 people, among whom 842 deaths were recorded during follow-up. After adjustment for age, sex, body mass index, smoking, alcohol consumption, education, and socioeconomic position, each one-standard-deviation increase in NoiseClock, which represents the typical spread of its values within the group, corresponded to an approximately 7% higher risk of death (95% confidence interval: 2% to 13%).
The age-related signal in a blood analysis may reflect several processes: changes in the mean levels of methylation marks, shifts in immune cell proportions, and increasing variation in methylation levels. This classification allows researchers to examine these processes separately.