(661.7) Systematic Integration of Epigenomic Landscapes in Human and Mouse Blood Cells to Predict Activity and Targets of Regulatory Elements

Poster Board Number: A242

Ross Hardison (The Pennsylvania State University), Guanjue Xiang (Dana Farber Cancer Institute), Camden Jansen (The Pennsylvania State University), Cheryl Keller (The Pennsylvania State University), Belinda Giardine (The Pennsylvania State University), April Cockburn (The Pennsylvania State University), Xi He (The Pennsylvania State University), Michael Sauria (Johns Hopkins University), Kathryn Weaver (Johns Hopkins University), Qunhua Li (The Pennsylvania State University), Yu Zhang (The Pennsylvania State University)

Thousands of genome-wide epigenetic (epigenomic) datasets determined in the past decade have generated maps of chromatin accessibility, histone modifications, and binding by transcription factors and structural proteins in a large number of cell types in humans and model species. Such maps have revealed many informative associations between groups of epigenetic features and different classes of cis-regulatory elements (CREs), and they enable a prediction of the locations of candidate CREs (cCREs) in many genomes. The actuation (opening of chromatin) and epigenetic state (combination of features such as histone modifications) acquired at specific cCREs has been associated with specific gene activation. However, despite the large number and rich variety of available epigenomic datasets, it is difficult for researchers to effectively utilize all the data relevant to their projects. Furthermore, comprehensive rules and catalogs have not been established for which CREs regulate specific target genes and how. Integrative analysis can help meet these needs, and our VISION project provides a ValIdated Systematic IntegratiONof epigenomic data in hematopoiesis in human and mouse blood cells. After careful curation and normalization of epigenomic data from many sources, we systematically integrated these data to generate a comprehensive view of the regulatory landscape (epigenetic states) and predicted cCREs in differentiating hematopoietic cell types jointly in both mouse and human. Multiple machine learning approaches reveal target genes predicted to be regulated by each element in different cell types. These approaches include predicting the regulatory impact of each element on targets by regression models, finding linkages between distinctive clusters of elements and potential target genes from self-organizing maps, and visual overlays of chromatin interaction maps. Directed mutation and expression analysis around selected genes supports the utility of the assignments. These catalogs of highly annotated regulatory elements are useful for many applications, including interpretation of the phenotypic impact of genetic variants in noncoding genomic regions. Resources from the VISION project are available at our website usevision.org.

Supported by NIH grants R24DK106766, R01DK054937, R01GM109453, R01GM121613.



The cCREs predicted to regulate the Alas2 gene in erythroid cells (gray arcs), with estimates of regulatory impact (light blue), confirmation of links by chromatin interaction data (purple), epigenetic states across hematopoietic cell types, and topologically associated domain (bottom green bar).