Ontology-based annotations and semantic relations in large-scale (epi)genomics data.

Area of Science:

• Bioinformatics
• Genomics
• Data Science

Background:

• Public biological data repositories house vast experimental data, largely untapped for combined analysis.
• Integrating diverse datasets requires robust sample annotation and inter-experiment relation capabilities.

Purpose of the Study:

• To semantically annotate Gene Expression Omnibus samples metadata using biomedical ontologies.
• To enable quantitative measurement of semantic similarity between samples for dataset aggregation.
• To demonstrate the utility of semantic annotation for identifying homogeneous experimental groups.

Main Methods:

• Applied semantic annotation to thousands of chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) samples using biomedical ontologies.
• Quantitatively measured semantic similarity between samples.
• Compared ontology-based tools (UMLS vs. topic-specific) for annotation and similarity computation.
• Identified and expanded datasets of Myc transcription factor ChIP-seq samples.

Main Results:

• Topic-specific ontologies outperformed Unified Medical Language System tools in annotation and semantic similarity measures.
• Successfully identified semantically homogeneous groups of Myc ChIP-seq samples.
• Expanded the Myc dataset with coherent epigenetic samples, validating semantic coherence with experimental data.

Conclusions:

• Semantic annotation of biological metadata is a powerful approach for integrating large-scale datasets.
• Topic-specific ontologies provide superior performance for biological data analysis compared to general systems.
• This method facilitates the generation of comprehensive datasets for studying specific biological targets and pathways.