Targeted recruitment of histone modifications in humans predicted by genomic sequences

Histone modifications are important epigenetic regulators and play a critical role in development. The targeting mechanism for histone modifications is complex and still incompletely understood. Here we applied a computational approach to predict genome-scale histone modification targets in humans by the genomic DNA sequences using a set of recent ChIP-seq data. We found that a number of histone modification marks could be predicted with high accuracy. On the other hand, the impact of DNA sequences for each mark is intrinsically different dependent upon the target- and tissue-specificity. Diverse patterns are associated with different repetitive elements. Unexpectedly, we found that non-overlapping, functionally opposite histone modification marks could share similar sequence features. We propose that these marks may target a common set of loci but are mutually exclusive and that the competition may be important for developmental control. Taken together, we show that our computational approach has provided new insights into the targeting mechanism of histone modifications.