Uncertainty Quantification in Epigenetic Clocks via Conformalized Quantile Regression.

Area of Science:

• Epigenetics
• Biostatistics
• Computational Biology

Background:

• DNA methylation (DNAm) influences biological age, with epigenetic clocks predicting age from DNAm levels.
• Epigenetic age acceleration may indicate health status and disease risk.
• Current epigenetic clocks lack uncertainty quantification, crucial for clinical applications.

Purpose of the Study:

• To develop a novel pipeline for training epigenetic clocks that quantifies uncertainty.
• To reveal population heterogeneity and construct accurate prediction intervals for biological age.
• To improve the understanding of epigenetic age beyond mean predictions.

Main Methods:

• Integration of high-dimensional quantile regression and conformal prediction.
• Training epigenetic clocks using DNAm data from 728 blood samples across 11 datasets in children.
• Developing adaptive prediction intervals that account for individual variability.

Main Results:

• The proposed method produces narrower prediction intervals compared to conventional mean regression-based clocks.
• Demonstrated improved statistical efficiency over existing epigenetic clock training pipelines.
• Revealed synchronized age acceleration patterns and cellular evolutionary heterogeneity in children and adolescents.

Conclusions:

• Conformalized high-dimensional quantile regression effectively produces valid prediction intervals and uncovers population heterogeneity.
• The methodology enhances understanding of epigenetic age, applicable beyond childhood.
• This toolbox offers insights for epigenetic interventions in age-related diseases.