Predicting genome-wide DNA methylation using methylation marks, genomic position, and DNA regulatory elements

Area of Science:

• Epigenetics
• Genomics
• Computational Biology

Background:

• Epigenome-wide association studies (EWAS) identify CpG sites linked to phenotypes using genome-wide DNA methylation profiles.
• Current computational methods predict average methylation and are limited to specific genomic regions.

Purpose of the Study:

• To develop a computational approach for predicting DNA methylation levels at single CpG site resolution.
• To identify genomic features that influence DNA methylation patterns.

Main Methods:

• Characterized genome-wide DNA methylation patterns to understand CpG site correlations.
• Developed a random forest classifier using features like neighboring CpG methylation, genomic distance, and regulatory elements (ENCODE).

Main Results:

• Achieved 92% prediction accuracy for genome-wide methylation at single-CpG-site precision.
• Accuracy increased to 98% for CpG sites within CpG islands (CGIs).
• Identified neighboring CpG methylation, CGIs, and regulatory elements as key predictive features.

Conclusions:

• Developed a highly accurate classifier for predicting DNA methylation at CpG site resolution.
• Identified genomic features interacting with DNA methylation, suggesting regulatory mechanisms.
• Linked diverse epigenetic processes through DNA methylation pattern analysis.