-Epigenomics is crucial to understand biological mechanisms beyond genome DNA. To better represent epigenomic knowledge and support data integration, we developed a prototype Epigenome Ontology (EGO). EGO top level hierarchy and design pattern are provided with a use case illustration. EGO is proposed to be used for statistically analyzing enriched epigenomic features based on given sequence data input using statistical methods.
The majority of eukaryotic genomes such as those of the human and mouse is noncoding. In eukaryote genomes, basic biological functions such as gene expression are affected by many regulatory elements located outside the coding region of the genome. Unlike the genome, which is largely static across tissues and cells within an individual, the epigenome is cell type specific and can be dynamically altered by environmental conditions. Better epigenomic understanding is critical for uncovering biological mechanisms and disease etiology.
improved in the past decade thanks to the efforts of several large international consortia, e.g., the Encyclopedia of DNA Elements (ENCODE; https://www.encodeproject.org/). As more and more cells and cell lines are being profiled, combined with more factors being studied, it becomes difficult to track all the experiments and resulting knowledge. In particular, many of the experiments are related due to the fact that either the cell types, or the experimental factors are related or both. Thus it is inadvisable to treat these data as independent.
For better interpretation, the subtle and complicated relationships among these experiments should be more fully considered to achieve a better interpretation. Specifically, a well-defined ontology system that handles complex relationships within a rigorous framework and offers annotation at various levels of granularity would be particularly effective. Towards such goals, we developed a new ontology named Epigenome Ontology (EGO).
A. EGO ontology development
(http://obofoundry.org/). Existing terms from other ontologies were imported to EGO using OntoFox (http://ontofox.hegroup.org). New terms were added and edited using the Protégé OWL editor.
continuant
(BFO) entity (BFO) occurrent (BFO) realizable entity (BFO) immaterial entity (BFO) material entity (BFO) process (BFO) disposition (BFO) site (BFO) cell (CL) genomic location (EGO) cell in vitro disease (DOID) molecular function (GO) human genomic location (EGO) cultured cell protein binding (GO)
human chromosome I6 location (EGO)
in U2OS cell (EGO) position of human Chr I6 (EGO) cell line cell (CLO)
U2OS cell (CLO) epigenome (EGO)
eukayotic epigenome (EGO) human epigenome (EGO) organism ‘has part’ (protein ‘has DNA template’ gene) biological process (GO) planned process (OBI) cellular process (GO) assay (OBI) chromatin organization
(GO) sequencing assay (OBI)
histone modification
(GO)
Current prototype EGO includes >600 terms. EGO is aligned with the BFO (http://ifomis.uni-saarland.de/bfo/) (Fig. 1). EGO also imports many terms and relations from existing ontologies (Fig. 1). EGO-specific terms focus on terms in the domain of epigenome. EGO defines ‘epigenome’ as a material entity that is made up of chemical compounds and proteins that can attach to DNA and direct various actions (e.g., turning genes on or off) (https://www.genome.gov/27532724). Since EGO targets genetic intervals at base-pair resolution, we will represent different positions of human and chromosomes and the events that occur at these positions. mouse B. EGO ontology design pattern with example
high quality, genome-wide profiling data and label every base
with observed “events” denoted as “EGO terms” such as
ChIPseq read coverage for a transcription factor (TF) in a specific
cell line. Fig. 2 illustrates an example design of how EGO
represents ChiP-seq read results for human Forkhead TF
FOXM1 binding in the cell line cell U2OS, which occurs at
different chromosomal positions of the cell line genome DNA
[
factor has_gene_template gene FOXM1 (OGG) function of
pbr(ioOncdBeinIs)gs ouhtapsut prcoot(emGinOp-lD)exNA inhpaust is_realized_in transcription molecular factor binding is a function (GO) (GO) functioisna(BFO) is a FOXM1 binding (EGO) is a gene (OGG)
ChiP-seq assay (OBI)
has is about specified output
ChiP-seq is a data item reads (OBI) reads on position isfuanctFioOinnXUi(nME2c1GOebOSllin)(csdehilnlogrtcub(sptin)hodosirintticgounat)t PisCohaosrfs2Ihi63tue5(mE9(Ba7GFn6OO3)7) paorft gseprietnreogeismi(oPBanoLFtoeKOfr1),
Ul2i(nCOeLScOec)lelll is a ceclell(lClinLeO) is a cell (CL)
With the EGO support, we will be able to leverage dense,
genome-wide annotations of various types to facilitate
enrichment analysis of “EGO terms” for any give genomic
region(s), similar to the GO term enrichment analyses or gene
set enrichment analysis (GSEA). As an illustration, suppose a
researcher conducted ATAC-seq experiments [
Such an enrichment analysis enables researchers to learn the key properties of a set of genomic regions, collectively, avoid the distraction of noise and diversity that are found common place in set of regions identified through highthroughput experiments. A key advantage of using an ontology system is that annotation is provided at different level of granularity, beyond individual dataset level. For example, instead of querying individual TFs in which many are highly related, EGO analysis enables query against TF families which provide concise and clear interpretation. Similarly, often times it is highly desirable to compare the binding patters of a factor across a handful of cell type classes over many closely-related individual cell types.
the domain of epigenome. With EGO, we are hoping to shed light on the less understood, “dark matter” regions of the genomes while producing insights, enhancing interpretation, and generating new hypotheses for genomic regions of interest.