New AI Model Combines DNA Sequence and Methylation for Epigenetic Age Prediction
Researchers have developed a novel machine learning framework that integrates DNA sequence context with methylation graph structure for epigenetic age prediction. The model uses a gated modulation mechanism to adaptively scale methylation signals based on sequence-determined biological relevance before graph convolution. Tested on 3,707 blood methylation samples, the approach outperforms existing methods that treat sequence and graph separately. This unified framework could improve accuracy in aging research and age-related disease studies.
Key facts
- The model integrates eight-dimensional DNA sequence statistical features with co-methylation graph structure.
- A lightweight gated modulation mechanism adaptively scales methylation signals based on sequence context.
- Evaluated on 3,707 blood methylation samples.
- No existing method jointly models co-methylation graph structure and site-specific DNA sequence context.
- Epigenetic clocks based on DNA methylation estimate biological age.
- Applications include aging research, age-related disease studies, and longevity science.
- Previous approaches include penalised linear regression, deep feedforward networks, residual architectures, and graph neural networks.
- The framework is described in arXiv:2605.10541v1.
Entities
—