Automatic detection of genomic regions with informative epigenetic patterns

• Published in: BMC genomics Vol. 19; no. 1; p. 847
• Main Authors: Pazos, Florencio, Garcia-Moreno, Adrian, Oliveros, Juan C
• Format: Journal Article
• Language: English
• Published: England BioMed Central Ltd 28.11.2018; BioMed Central; BMC
• Subjects: Analysis; Automation; Bioinformatics; Brain; Brain - metabolism; Cancer; Classification; Computational Biology - methods; Data collection; Databases, Genetic; Datasets; Deoxyribonucleic acid; Developmental plasticity; Developmental stages; DNA; DNA sequencing; Epigenesis, Genetic; Epigenetic inheritance; Epigenetics; Epigenomics; Fetus - metabolism; Fetuses; Gene expression; Gene transcription regulation; Genome, Human; Genomes; Genomics; Humans; Information retrieval; Methods; Neoplasms - genetics; Nucleotide sequence; Ontology; Organs; Phenotypic plasticity; Proteins; Repositories; Tissues; Epigenetics; Gene transcription regulation; Epigenomics
• DOI: 10.1186/s12864-018-5286-5
• ISSN: 1471-2164

Epigenetic phenomena are crucial for explaining the phenotypic plasticity seen in the cells of different tissues, developmental stages and diseases, all holding the same DNA sequence. As technology is allowing to retrieve epigenetic information in a genome-wide fashion, massive epigenomic datasets are being accumulated in public repositories. New approaches are required to mine those data to extract useful knowledge. We present here an automatic approach for detecting genomic regions with epigenetic variation patterns across samples related to a grouping of these samples, as a way of detecting regions functionally associated to the phenomenon behind the classification. We show that the regions automatically detected by the method in the whole human genome associated to three different classifications of a set of epigenomes (cancer vs. healthy, brain vs. other organs, and fetal vs. adult tissues) are enriched in genes associated to these processes. The method is fully automatic and can exhaustively scan the whole human genome at any resolution using large collections of epigenomes as input, although it also produces good results with small datasets. Consequently, it will be valuable for obtaining functional information from the incoming epigenomic information as it continues to accumulate.