Novel AI Framework Integrates Sequence and Graph Data for Precise Epigenetic Age Prediction
Researchers have developed a unified sequence-graph integration framework to enhance epigenetic age prediction using DNA methylation data. Addressing a gap in existing machine learning models, this method jointly models co-methylation graph structures and site-specific DNA sequence contexts. The approach utilizes eight-dimensional DNA sequence statistical features via a lightweight gated modulation mechanism, which adaptively scales methylation signals based on biological relevance before graph convolution. Evaluated on 3,707 blood methylation samples, the model achieved a test Mean Absolute Error (MAE) of 3.149 years, representing a 12.8% improvement over the strongest graph-based baseline. Notably, the study found that handcrafted, biologically informed statistical features outperformed CNN-based sequence encoding, suggesting their superior effectiveness in this specific data regime. Interpretability analysis highlighted CpG density and local adenine frequency as key features with age-dependent importance shifts, aligning with known mechanisms of age-related hypermethylation. This advancement offers significant potential for aging research, longevity science, and the study of age-related diseases by providing more accurate biological age estimates through improved computational modeling.
Wire timeline
Novel AI Framework Integrates Sequence and Graph Data for Precise Epigenetic Age Prediction
Researchers have developed a unified sequence-graph integration framework to enhance epigenetic age prediction using DNA methylation data. Addressing a gap in existing machine learning models, this method jointly models co-methylation graph structures and site-specific DNA sequence contexts. The approach utilizes eight-dimensional DNA sequence statistical features via a lightweight gated modulation mechanism, which adaptively scales methylation signals based on biological relevance before graph convolution. Evaluated on 3,707 blood methylation samples, the model achieved a test Mean Absolute Error (MAE) of 3.149 years, representing a 12.8% improvement over the strongest graph-based baseline. Notably, the study found that handcrafted, biologically informed statistical features outperformed CNN-based sequence encoding, suggesting their superior effectiveness in this specific data regime. Interpretability analysis highlighted CpG density and local adenine frequency as key features with age-dependent importance shifts, aligning with known mechanisms of age-related hypermethylation. This advancement offers significant potential for aging research, longevity science, and the study of age-related diseases by providing more accurate biological age estimates through improved computational modeling.
cs.AI updates on arXiv.org