Methylation Prediction in Human Genome
 
Zhang Lab, Cold Spring Harbor Laboratory
 

Background
Paper
Supplemental
Prediction
Software
License
Contact


 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BACKGROUND

Epigenetic effects in mammals depend largely on heritable cytosine methylation patterns, but little is known of target sequence specificity for de novo cytosine methylation. We have identified discriminating sequence features enriched in methylated or unmethylated regions, and their potential relations to known functional elements. We describe a computational pattern recognition method that can predict the genomic DNA methylation profiles in the human adult brain. The algorithm computes the methylation propensity for an 800 bp sequence region centered around a CpG dinucleotide based on specific sequence features within the region. Our program (called HDMFinder) has a prediction accuracy of 86%, as validated with CpGs regions for which methylation status have been experimentally determined. Using HDMFinder, we are able to depict the entire genomic methylation patterns for all 22 human autosomes.