Epigenomic variation between species, tissues, populations and individuals. A. Kundaje1,2, W. Meuleman1,2, J. Wang1,2, N. Kumar1, S. Kyriazopoulou-Panagiotopoulou3, M. Kasowski4, M. Snyder4, M. Kellis1,2 1) Computer Science Dept., Massachusetts Institute of Technology, Cambridge, MA. 02139; 2) Broad Institute of MIT and Harvard, Cambridge, MA 02139; 3) Computer Science Dept., Stanford University, Stanford, CA, 94305; 4) Dept. of Genetics, Stanford University, Stanford, CA, 94305.

   In multicellular organisms, epigenetic information is a key enabler of dynamic regulatory regions shaping the identity of each cell. This information is encoded in distinct combinations of epigenetic modifications defining chromatin states specific to different types of functional elements such as promoters, enhancers, transcribed elements and repressed domains. Here, we provide the first systematic analysis of chromatin state dynamics across 3 axes of variation: inter-species variation, inter-cell type variation in the same species, and inter-individual variation for the same cell type. First, we used genome-wide maps of histone modifications from the mod/ENCODE consortia across human, fly, and worm to learn unified chromatin state models and study the similarities and differences in the specific mark combinations that define different types of functional elements. We found remarkable conservation of the chromatin code across the 3 species, for active enhancers, promoter, and transcribed regions, illustrating a universality of chromatin states across the animal kingdom. However, we also find a notable exception in the marking of repressed regions, where repressed Polycomb and heterochromatin domains that are distinct in fly and human appear to co-associate in worm, indicating potentially different mechanisms of repression. Second, we learned robust chromatin state maps in 90 human cell types using data from the Roadmap Epigenomics consortium that enabled us to construct detailed lineage trees relating the different cell-types. We developed novel predictive models to link distal enhancers to their target genes that revealed extensive cell-type specific activity of both enhancer and promoter regions associated with distinct cell-type specific regulatory programs. Finally, we learned chromatin state maps jointly across lymphoblastoid cell lines from 19 individuals spanning diverse ancestry. We find that enhancer regions are the most variable functional elements across individuals, showing significant switching between active, weak, bivalent/poised, and repressed states. In contrast, promoter and transcribed states showed negligible variation across individuals and populations, suggesting that the observed enhancer variation is inconsequential or potentially buffered by enhancer redundancy. Together, these analyses provide key insights into the conservation and variation of chromatin regulation and dynamics across organisms, cell-types and individuals.

You may contact the first author (during and after the meeting) at