Time is encoded by methylation changes at clustered CpG sites.

Area of Science:

• Genetics
• Epigenetics
• Computational Biology

Background:

• DNA methylation changes with age, enabling biological and chronological age estimation.
• The precise mechanisms driving these age-dependent DNA methylation alterations are not fully understood.
• Current epigenetic clocks have limitations in accuracy and the underlying cellular processes.

Purpose of the Study:

• To investigate the regional and stochastic mechanisms of age-dependent DNA methylation.
• To develop a novel, highly accurate method for chronological age prediction using DNA methylation.
• To explore the cellular basis of age encoding within DNA methylation patterns.

Main Methods:

• Ultra-deep sequencing of DNA methylation across >300 blood samples.
• Deep learning analysis of single-molecule DNA methylation patterns in specific genomic loci.
• Validation of age prediction accuracy on independent human blood samples.

Main Results:

• Age-dependent DNA methylation changes occur regionally at adjacent CpG sites, stochastically or in blocks.
• Deep learning models achieved highly accurate chronological age prediction (1.46-1.7 years median error) from DNA methylation.
• Chronological age could be inferred from as few as 50 DNA molecules, indicating cellular-level age encoding.
• Factors like gender, BMI, and smoking did not impact chronological age inference.

Conclusions:

• Clustered DNA methylation changes provide fundamental insights into cellular and tissue time measurement.
• The developed deep learning approach significantly enhances the accuracy of epigenetic clocks.
• This research has potential applications in medical diagnostics and forensic science for age determination.