PheneBank

PheneBank extracts and validates human phenotype–disease associations from Medline using machine learning–based text mining to integrate literature-derived phenotypic annotations into ontologies such as the Human Phenotype Ontology (HPO).

Key Features:

• Text Mining Capabilities: Employs machine learning algorithms to identify phenotype concepts in Medline using an expert-annotated rare disease corpus from the PMC Text Mining subset.
• Integration with Existing Ontologies: Produces literature-based phenotype annotations that can be mapped to the Human Phenotype Ontology (HPO) and other coding systems.
• Validation Against Gold Standards: Evaluates performance using a gold-standard corpus of rare disease sentences and cross-references extracted associations with the Monarch Initiative.

Scientific Applications:

• Phenotype–disease association discovery: Automates extraction of candidate phenotype–disease links from literature to support discovery of novel associations.
• Phenotype database curation: Supplements and expands phenotype databases by providing literature-derived annotations for integration into ontologies and coding systems.
• Rare disease research: Enhances aggregation and validation of detailed phenotypic information pertinent to rare disease analyses and hypothesis generation.

Methodology:

Machine learning–based text mining models trained on expert-annotated rare disease datasets from the PMC Text Mining subset and Medline perform contextual analysis to identify phenotype mentions; performance is evaluated against a gold-standard corpus and cross-referenced with the Monarch Initiative.