Time is encoded by methylation changes at clustered CpG sites

Bracha-Lea Ochana,Daniel Nudelman,Daniel Cohen,Ayelet Peretz,Sheina Piyanzin,Ofer Gal,Amit Horn,Netanel Loyfer,Miri Varshavsky,Ron Raisch,Ilona Shapiro,Yechiel Friedlander,Benjamin Glaser,Hagit Hochner,Yuval Dor,Tommy Kaplan,Ruth Shemer

DOI: https://doi.org/10.1101/2024.12.03.626674

2024-12-05

Abstract:Age-dependent changes in DNA methylation allow chronological and biological age inference, but the underlying mechanisms remain unclear. Using ultra-deep sequencing of >300 blood samples from healthy individuals, we show that age-dependent DNA methylation changes are regional and occur at multiple adjacent CpG sites, either stochastically or in a coordinated block-like manner. Deep learning analysis of single-molecule patterns in two genomic loci achieved accurate age prediction with a median error of 1.46-1.7 years on held-out human blood samples, dramatically improving current epigenetic clocks. Factors such as gender, BMI, smoking and other measures of biological aging do not affect chronological age inference. Longitudinal 10-year samples revealed that early deviations from epigenetic age are maintained throughout life and subsequent changes faithfully record time. Lastly, the model inferred chronological age from as few as 50 DNA molecules, suggesting that age is encoded by individual cells. Overall, DNA methylation changes in clustered CpG sites illuminate the principles of time measurement by cells and tissues, and facilitate medical and forensic applications.

Biology

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict an individual's physiological age more accurately by analyzing DNA methylation changes. Specifically, researchers are concerned with how the methylation patterns of DNA in specific regions (such as multiple adjacent CpG sites) change as people age, and explore whether these changes can be used to construct a biological clock model that is more precise than existing methods. The main contributions of the paper are as follows:

1. **Regional DNA methylation changes**: The study found that age - related DNA methylation changes do not occur in isolation at a single CpG site, but occur in a regional manner at multiple adjacent CpG sites. These changes may be random or coordinated block - like changes.

2. **Analysis at the single - molecule level**: Through ultra - deep sequencing technology, researchers were able to analyze the methylation patterns on a single DNA molecule, thus capturing more information, which provides a higher resolution than traditional array - based methods.

3. **Deep - learning model**: The researchers developed a deep neural network model (MAgeNet), which can learn from the combined methylation patterns of multiple CpG sites and be used to predict an individual's physiological age. The median error of this model on the test set is only 1.46 - 1.7 years, which is significantly better than the existing epigenetic clocks.

4. **The influence of environmental factors**: The study also explored the influence of environmental factors such as gender, BMI, and smoking on age prediction. The results show that these factors do not affect the accuracy of age prediction based on DNA methylation.

5. **The requirement for the minimum number of cells**: The study further explored the minimum number of cells required for age prediction, which is of great significance in fields such as forensic applications. The results show that even with a small number of DNA molecules (such as 50), relatively accurate age prediction can be achieved.

In conclusion, through in - depth analysis of the change patterns of DNA methylation, this paper proposes a new and more accurate age - prediction method, which not only helps to understand the biological mechanisms of aging, but also has broad application prospects in fields such as clinical diagnosis and forensic identification.