PhenoMiner: from text to a database of phenotypes associated with OMIM diseases

Area of Science:

• Computational biology
• Bioinformatics
• Genetics

Background:

• Manual curation of scientific literature for phenotype data is time-consuming and faces challenges due to diverse human expressivity.
• Existing databases like OMIM rely on manual efforts, limiting scalability and comprehensive data integration.

Purpose of the Study:

• To introduce PhenoMiner, a novel automated approach for extracting scientific and clinical phenotypes from literature.
• To identify and harmonize phenotype descriptions and discover phenotype-disease relationships.
• To build a comprehensive database of phenotype-disorder associations.

Main Methods:

• PhenoMiner utilizes full parsing and conceptual analysis for automated phenotype extraction.
• Apriori association mining is employed to identify relationships between phenotypes and human diseases.
• The approach was applied to the BMC open access collection.

Main Results:

• PhenoMiner identified 13,636 phenotype candidates and 28,155 phenotype-disorder hypotheses.
• These hypotheses covered 4898 phenotypes and 1659 Mendelian disorders.
• Analysis confirmed semantic distribution against ontologies, overlap with Human Phenotype Ontology (HP), and strong associations with known disease-gene pairs.

Conclusions:

• PhenoMiner offers an effective automated solution for phenotype extraction and phenotype-disease relationship discovery.
• The tool enhances the quality and scalability of phenotype data curation for genetic research.
• The generated data provides valuable resources for understanding human expressivity and genetic disorders.