In a recent study published in Aging, researchers identified deoxyribonucleic acid (DNA) methylation entropy as a biological marker for age predictions.
Advancing age gradually reduces epigenetic information, which could inform future treatments for age-related diseases. Entropy denotes the extent of epigenetic variations in methylation. Using entropy as a metric may enhance the utility of epigenetic clocks and refine them for clinical use.
Image Credit: Grustock/Shutterstock.com
Introduction
Somatic cells differentiate into various types through epigenetic modifications. DNA methylation, particularly at cytosine-guanine (CpG) dinucleotides, is a key epigenetic modification. Epigenetic clocks estimate chronological age from age-related changes in methylation. These clocks can produce biomarkers to predict lifespan.
Studies in mice suggest that the patterns of DNA methylation accumulate stochastically over time, potentially serving as a molecular biomarker to predict chronological age. However, it is unclear whether this approach can estimate the chronological age of humans.
Moreover, the scientific community needs to compare the performance of models using DNA methylation for age estimation to traditional epigenetic analysis for clinical validation.
About the study
The present study researchers explored the relationship between DNA methylation and chronological age. In particular, they investigated whether the accumulation of epigenetic mutations changes reproducibly with age.
The team obtained buccal swab samples from 100 participants aged between 7.2 and 84 years. Deoxyribonucleic acid (DNA) extracted from the samples underwent polymerase chain reaction (PCR) amplification to generate age-related methylation profiles by targeted bisulfite sequencing (TBS).
TBS measured DNA methylation levels across 3000 loci to compute single cytosine-based and read-based metrics and enumerated patterns of DNA methylation at each locus.
Researchers assessed the distribution of methylation patterns to assess age-associated entropy changes at particular loci. They compared these changes to those related to individual site methylation.
They achieved an average of 293 genetic reads per swab sample for methylation analysis. Penalized regressions evaluated the capability of the metrics to estimate the chronological age.
In addition, the researchers investigated the Cellular Heterogeneity-Adjusted cLonal Methylation (CHALM) to assess region-specific and read-level methylations after classifying the reads as unmethylated or methylated based on methylcytosine presence.
They used the Shannon entropy formula to obtain DNA methylation entropy values for each genetic locus, including four sites of CpG dinucleotides in every region.
The team focused on two loci showing extreme correlations with entropy and age-chr15:51681883-51681783 and chr2:101001739-101001859. Elastic net and neural network regressions predicted age using average methylation, CHALM, and entropy.
The researchers evaluated model performance using leave-one-out cross-validation (LOOCV), Pearson correlation coefficients between the projected and actual age, and the mean absolute error (MAE) of the projected age. They compared the findings with Horvath's epigenetic clock to assess effectiveness.
Results
The study suggests that epigenetic clocks based on DNA methylation entropy can predict chronological age with a high degree of accuracy, comparable to traditional methods that rely on the methylation levels of individual cytosines.
The models achieved correlation coefficients above 0.90, indicating strong predictive capability. Age-related changes in mean methylation showed strong correlations with CHALM values.
The findings suggest that the accumulation of epimutations, as quantified by Shannon's entropy, changes reproducibly with age. Thus, the diversity of methylation patterns is a crucial factor in aging.
At locus chr15:51681883-51681783, the team observed positive correlations of entropy with age. They obtained Pearson correlation coefficient values of 0.82 between average methylation and age and 0.79 between entropy and age. Young samples showed hypomethylated reads, whereas old samples showed diverse methylation patterns.
Locus chr2:101001739-101001859 showed a positive correlation coefficient of 0.73 between average DNA methylation and individual age but a negative correlation of -0.63 between entropy and age. Young samples showed discrete methylation patterns, whereas old samples primarily exhibited fully methylated DNA patterns.
Pearson correlation coefficients were in the 0.79-0.93 range across models, with a mean absolute error of four to eight years. Elastic net regressions consistently yielded models with high correlation and low MAE. Neural network regressions showed variability, with some models performing better than others.
Models using entropy with all three metrics performed the best, demonstrating correlation coefficients exceeding 0.90 and an average error of only five years. Comparison with Horvath's clock showed similar performance for mean methylation-based models, with a correlation coefficient of 0.85 and an MAE of 7.1 years.
Correlations with age were similar across metrics, but entropy showed complex patterns. Entropy can increase or decrease with age, independent of methylation levels.
If a locus is initially hypo- or hypermethylated, methylation patterns diversify, and entropy improves with age. Conversely, sites starting with high entropy early in life lose diversity and entropy with age.
The findings suggest that DNA methylation entropy could indicate chronological age from epigenetic data, supporting the Information Theory of Aging. Measuring the entropy of DNA methylation profiles may be a more informative aging biomarker than conventional approaches based on individual site methylation levels.
Combining methylation entropy with mean-based DNA methylation metrics yields the most accurate results. Future studies could elucidate the mechanisms underlying the relationship between aging and entropy in various tissues with diverse methylation patterns to increase the applicability of the DNA methylation-based approach for chronological age estimation.
DNA Computing Circuits Move Closer to Real-Time Function Inside Living Cells