Exploratory visualizations and statistical analysis of large, heterogeneous epigenetic datasets

Epigenetic marks, such as DNA methylation and histone modifications, are important regulatory mechanisms that allow a single genomic sequence to give rise to a complex multicellular organism. When studying mechanisms of epigenetic regulation, the analyses depend on the experimental technologies and the available data. Recent advancements in sequencing technologies allow for the efficient extraction of genome-wide maps of epigenetic marks. A number of large-scale mapping projects, such as ENCODE and IHEC, intensively produce data for different tissues and cell cultures. The increasing quantity of data highlights a major bottleneck in bioinformatic research, namely the lack of bioinformatic tools for analyzing these data. To date, there are bioinformatics tools for detailed (mostly visual) inspection of single genomic loci, allowing biologists to focus research on regions of interest. Also, efficient tools for manipulation and analysis of the data have been published, but often they require computer science abilities. Furthermore, the available tools provide solutions to only already well formulated biological questions. What is missing, in our opinion, are tools (or pipelines of tools) to explore the data interactively, in a process that would facilitate a trained biologist to recognize interesting aspects and pursue them further until concrete hypotheses are formulated. A possible solution stems from the best practices in the fields of information retrieval and exploratory search. In this thesis, I propose EpiExplorer, a paradigm for integration of state-of-the-art information retrieval methods and indexing structures, applied to offer instant interactive exploration of large epigenetic datasets. The algorithms we use are developed for semi-structured text data, but we apply them on bioinformatic data through clever textual mapping of biological properties. We demonstrate the power of EpiExplorer in a series of studies that address interesting biological problems. We also present in this manuscript EpiGRAPH, a bioinformatic software that we developed with colleagues. EpiGRAPH helps identify and model significant biological associations among epigenetic and genetic properties for sets of regions. Using EpiExplorer and EpiGRAPH, independently or in a pipeline, provides the bioinformatic community with access to large databases of annotations, allows for exploratory visualizations or statistical analysis and facilitates reproduction and sharing of results.