15 Jul Hebrew University: AI Model Accurately Predicts Age from Examining Just Two Gene Loci

The study was done by a team of researchers at the The Hebrew University-Hadassah Medical School, led by Bracha Ochana and Daniel Nudelman, under the supervision of Prof. Tommy Kaplan, Prof. Yuval Dor and Prof. Ruth Shemer.

Prof. Tommy Kaplan

MedicalResearch.com: What is the background for this study?

Response: DNA methylation is a key epigenetic modification that annotates the human genome. It is established during development and cellular differentiation, and is associated with maintenance of cell type identity and control of gene expression. Nonetheless, few regions in the human genome change with age and serves as a powerful biomarker for estimating chronological and biological age. However, most current epigenetic clocks rely on average methylation at individual CpG sites using array-based data, which overlook complex regional patterns across neighboring methylation sites. This study aimed to understand how time and age are encoded at the molecular and cellular level, and to develop a highly accurate age predictor, based on regional methylation dynamics.

Prof. Ruth Shemer

MedicalResearch.com: What types of cells were used in the study, ie keratinocytes, muscle cells etc?

Response: The primary tissue used in this study was peripheral blood from over 300 healthy human donors (18-78 years old). To further understand how the methylation changes are associated with changes in blood cell composition, we also sorted immune cell types including neutrophils, monocytes, B cells, and T cells. For forensics applications, we also tested the clock on urine and saliva samples.

Prof. Yuval Dor

MedicalResearch.com: Does this study relate at all to telomere length?Response:

Response: No, this study does not investigate or reference telomere length. It focuses entirely on DNA methylation changes at few genomics regions, each covering multiple clustered DNA methylation sites, where methylation changes are indicative of chronological age, independently of telomere biology.

MedicalResearch.com: What are the main findings?

Response:  – A single-molecule analysis using DNA sequencing, revealed that age-related methylation changes often occur regionally across multiple neighboring methylation sites, either in a stochastic or in a block-like manner.
– A deep neural network model, called MAgeNet, was trained on methylation patterns from two specific genomic loci (ELOVL2 and C1orf132) and was able to predict chronological age (of held-out test-set donors) at a median accuracy of 1.36 years (for individuals under 50).
– These predictions are robust to sex, smoking, BMI, and biological age markers, and accurate even from as few as 50 cells or at low-depth sequencing.
– Longitudinal sampling of healthy donors at the age of 32 and 42, shows that early deviations from predicted age persist over time, suggesting that as we age, methylation changes faithfully encode the passage of time. 

MedicalResearch.com: What should readers take away from your report?

Response: While we don’t fully understand how cells and tissues age, it was shown that time leaves a molecular footprint in our DNA. By examining how methylation patterns evolve at a regional manner, across neighboring sites, this study reveals that elapsed time is reliably encoded at the single-molecule level in human cells. The resulting model, MAgeNet, enables ultra-accurate age prediction with applications in medicine, aging research, and forensic science. 

MedicalResearch.com: What recommendations do you have for future research as a results of this study?

 – Investigate the molecular mechanisms driving regional age-dependent changes, across multiple sites.

– Study the functional and clinical role of these age-affected regions in the human genome

– Single-molecule data allows disentanglement of cell-type-specific age-related changes, compared to tissue-level dynamics  (e.g. alteration in cell-type composition).

– Extend this framework to other tissues and cell types to enable broad clinical and forensic use.

– Explore how early-life events or genetic factors influence long-term deviations in methylation-based age predictions.

– Develop similar clocks for predicting the biological age

MedicalResearch.com: Is there anything else you would like to add? Any disclosures?

Response: The authors declare no competing financial interests. All data and code (MAgeNet) are openly available via GitHub: https://github.com/danielnudel/MAgeNet. The study was supported by grants from the Israel Science Foundation, Horizon Europe, and the NIH, among others.

Citation:

Bracha-Lea Ochana, Daniel Nudelman, Daniel Cohen, Ayelet Peretz, Sheina Piyanzin, Ofer Gal Rosenberg, Amit Horn, Netanel Loyfer, Miri Varshavsky, Ron Raisch, Ilona Shapiro, Yechiel Friedlander, Hagit Hochner, Benjamin Glaser, Yuval Dor, Tommy Kaplan, Ruth Shemer,
Time is encoded by methylation changes at clustered CpG sites,
Cell Reports, 2025, 115958, ISSN 2211-1247,
https://doi.org/10.1016/j.celrep.2025.115958.

The information on MedicalResearch.com is provided for educational purposes only, and is in no way intended to diagnose, cure, or treat any medical or other condition. Some links are sponsored. Products are not warranted or endorsed.

Always seek the advice of your physician or other qualified health and ask your doctor any questions you may have regarding a medical condition. In addition to all other limitations and disclaimers in this agreement, service provider and its third party providers disclaim any liability or loss in connection with the content provided on this website.

Last Updated on July 15, 2025 by Marie Benz MD FAAD