A Global Genome Segmentation Method for Exploration of Epigenetic Patterns

Abstract

Current genome-wide ChIP-seq experiments on different epigenetic marks aim at unraveling the interplay between their regulation mechanisms. Published evaluation tools, however, allow testing for predefined hypotheses only. Here, we present a novel method for annotation-independent exploration of epigenetic data and their inter-correlation with other genome-wide features. Our method is based on a combinatorial genome segmentation solely using information on combinations of epigenetic marks. It does not require prior knowledge about the data (e.g. gene positions), but allows integrating the data in a straightforward manner. Thereby, it combines compression, clustering and visualization of the data in a single tool. Our method provides intuitive maps of epigenetic patterns across multiple levels of organization, e.g. of the co-occurrence of different epigenetic marks in different cell types. Thus, it facilitates the formulation of new hypotheses on the principles of epigenetic regulation. We apply our method to histone modification data on trimethylation of histone H3 at lysine 4, 9 and 27 in multi-potent and lineage-primed mouse cells, analyzing their combinatorial modification pattern as well as differentiation-related changes of single modifications. We demonstrate that our method is capable of reproducing recent findings of gene centered approaches, e.g. correlations between CpG-density and the analyzed histone modifications. Moreover, combining the clustered epigenetic data with information on the expression status of associated genes we classify differences in epigenetic status of e.g. house-keeping genes versus differentiation-related genes. Visualizing the distribution of modification states on the chromosomes, we discover strong patterns for chromosome X. For example, exclusively H3K9me3 marked segments are enriched, while poised and active states are rare. Hence, our method also provides new insights into chromosome-specific epigenetic patterns, opening up new questions how “epigenetic computation” is distributed over the genome in space and time.

On the Cooperation between Epigenetics and Transcription Factor Networks in the Specification of Tissue Stem Cells

“…DNMTs are repelled by H3K4me3 [11] and evidence has been provided that they are recruited by H3K27me3 [12].Based on such experimental findings, we have established a computational model of epigenetic regulation of transcription [13,14] and applied it in order to explain gene regulatory phenomena associated with stem cell differentiation and aging [15][16][17][18]. In parallel, we have developed data analysis tools that are based on self-organizing maps (SOMs) that enable a combined analysis of transcription and epigenetic data [19,20]. After applying these tools to data sets on developmental or tissue transformation processes, we observed that strong correlations between transcription and histone methylation, although existent for important steps of regulation, do not appear to be mandatory and apply only to a well-defined subset of genes.Here, we study such relations re-analyzing RNA-seq, H3K4me3, and H3K27me3 ChIP-seq as well as methyl-CpG binding domain protein (MBD)-seq data on cells of four different states of the specification and differentiation of mouse intestinal stem cells (ISCs) published by Kazakevych et al [21].…”

Computational Methods in Epigenetics

“…piClust works as follows: first of all, it determines the clustering parameters carrying out a k-dist analysis; then it clusters preprocessed and previously aligned reads; last, it scores and validates candidate piRNA clusters. Steiner et al [107] used an artificial neural network, more specifically a self-organizing map, to perform clustering, multidimensional scaling, and visualization of epigenetic patterns. It was believed that applying this tool to the epigenome would help in the understanding of the role of chromatin structure in gene expression regulation.…”

Analyzing Histone Modifications Using Tiled Binned Clustering and 3D Scatter Plots

“…Histone modifications are important for the development of the cells regulating the transcription states of genes and thereby their overall expression patterns. The evolution of certain histone modifications plays an important role during the differentiation of cells, e.g., from embryonic stem cells to embryonic fibroblasts or to neural progenitor cells [23,27,28].…”

“…In order to analyze this fate of histone modifications, one or more intermediate levels of data analysis are needed that go beyond the high-level statistical correlations and the low-level sequence-level analyses. Steiner et al [23] proposed a visualization based on self-organizing maps (SOMs) on an intermediate level.…”

“…Moreover, Zeckzer et al proposed tiled binned scatter plot matrices (TiBi-SPLOM) [28] and tiled binned 3D scatter plots (TiBi-3D) [27] fostering analyzing the fate of epigenetic marks during differentiation. While our previous approaches [23,27,28] provide effective and efficient visualizations of histone modification data supporting these intermediate level analyses, there is still room for improvements.…”