Evaluation of six blood-based age prediction models using DNA methylation analysis by pyrosequencing

Area of Science:

• Epigenetics
• Biomarkers
• Gerontology

Background:

• DNA methylation is a leading molecular biomarker for age prediction.
• Existing DNA methylation-based age prediction models, primarily using pyrosequencing on blood samples, show variable performance.
• Independent validation and intercomparison of these models are scarce.

Purpose of the Study:

• To evaluate and compare the performance of six established blood-based DNA methylation age prediction models.
• To assess model accuracy using correlation analysis, Mean Absolute Deviation (MAD), and Standard Error of the Estimate (SEE).
• To provide an independent validation of different age prediction models on a single cohort.

Main Methods:

• Pyrosequencing-based DNA methylation analysis was performed on 100 French blood samples (ages 19-65).
• Six models (Bekaert, Park, Thong, Weidner, Zbiec-Piekarska 1, Zbiec-Piekarska 2) were evaluated.
• Statistical analyses included correlation, MAD, and SEE calculations for each model.

Main Results:

• The Bekaert and Thong models demonstrated the best age prediction performance (MAD: 4.5-5.2, SEE: 6.8-7.2).
• The Zbiec-Piekarska 1 model showed moderate performance (MAD: 6.8, SEE: 9.2).
• Park, Weidner, and Zbiec-Piekarska 2 models exhibited lower predictive accuracy (MAD: 7.2-8.7, SEE: 9.2-10.3).

Conclusions:

• Independent evaluation revealed significant performance differences among blood-based DNA methylation age prediction models.
• The Bekaert and Thong models are recommended for accurate age prediction in similar cohorts.
• Systematic, independent validation of age prediction models on shared cohorts is crucial for reliable comparisons.