DNA methylation-based age estimation for adults and minors: considering sex-specific differences and non-linear...

Area of Science:

• Epigenetics and Aging Research
• Biomarker Discovery
• Computational Biology

Background:

• DNA methylation patterns exhibit age-related changes, making them potential biomarkers for estimating biological age.
• The relationship between DNA methylation and aging may be non-linear and influenced by sex.
• Previous models often overlook non-linear dynamics and sex-specific variations in methylation patterns.

Purpose of the Study:

• To comparatively evaluate linear and non-linear regression models for DNA methylation-based age prediction.
• To assess the impact of sex-specific versus unisex models on age estimation accuracy.
• To develop and validate an improved, non-linear, unisex DNA methylation age prediction model.

Main Methods:

• Analysis of DNA methylation patterns from buccal swabs of 230 individuals aged 1-88 years.
• Application of sequential replacement regression and 10-fold cross-validation on training and validation sets.
• Development of a non-linear, unisex model incorporating a 20-year age cut-off and specific methylation markers (EDARADD, KLF14, ELOVL2, FHL2, C1orf132, TRIM59).

Main Results:

• A non-linear, unisex model combining six methylation markers demonstrated robust age prediction accuracy.
• The model achieved a cross-validated Mean Absolute Deviation (MAD) of 4.680 years and Root Mean Square Error (RMSE) of 6.436 years in the training set.
• Sex-specific models showed improved accuracy in females, though not significantly in males, possibly due to sample size limitations.

Conclusions:

• The developed non-linear, unisex DNA methylation model provides a reliable method for age estimation.
• While age- and sex-adjustments did not universally enhance model performance, they may benefit other models and larger cohorts.
• The study highlights the potential of specific DNA methylation markers for accurate and accessible age prediction.