PheneBank: a literature-based database of phenotypes

Area of Science:

• Biomedical Informatics
• Computational Biology
• Medical Genetics

Background:

• Curated coding systems like the Human Phenotype Ontology (HPO) and disease-phenotype annotations are crucial for understanding genetic disorders.
• Discovering and integrating literature-based phenotypes into knowledge discovery processes remains a significant challenge in biomedical research.

Purpose of the Study:

• To develop and present PheneBank, a web portal for retrieving human phenotype-disease associations.
• To support the automated discovery of phenotypes from scientific literature and integrate them into existing knowledge frameworks.

Main Methods:

• Leveraging state-of-the-art machine learning for concept identification.
• Utilizing an expert-annotated rare disease corpus from the PMC Text Mining subset for training.
• Text mining the entire Medline database to extract phenotype-disease associations.

Main Results:

• PheneBank provides a comprehensive resource of text-mined human phenotype-disease associations.
• The system demonstrates effective concept identification and association extraction.
• Evaluation against gold-standard corpora and established initiatives confirms system accuracy.

Conclusions:

• PheneBank facilitates the discovery of literature-based phenotypes, enhancing biomedical knowledge.
• The portal integrates extracted associations, supporting research in rare diseases and human genetics.
• Accessible data and software promote further development in phenotype analysis and knowledge discovery.