KCCG Patient Archive

KCCG Patient Archive aggregates and annotates clinical and genomic phenotype data using Human Phenotype Ontology (HPO) concept recognition to develop disease models and support phenotype-driven analysis of sequence-variation data.

Key Features:

• Concept recognition and annotation: A corpus of 228 manually annotated abstracts using HPO concepts, harmonized by three curators, provides a reference standard for free-text annotation of human phenotypes.
• Standardized error analysis test suite: A test suite comprising 32 different types of test cases corresponding to 2,164 HPO concepts enables standardized evaluation of concept recognition errors.
• Evaluation of phenotype recognizers: Comparative evaluations have been performed for NCBO Annotator, OBO Annotator, and Bio-LarK CR against the annotated corpus and the standardized test suites.
• Disease model development for common diseases: Concept-recognition procedures analyze frequencies of HPO disease annotations across over five million PubMed abstracts with iterative optimization to improve precision and recall, yielding models for 3,145 common human diseases with 132,006 HPO annotations.
• Phenotypic overlap analysis: Analysis of phenotypic overlap among common diseases that share risk alleles and between Mendelian and common diseases linked by genomic location supports phenotype-driven analysis of next-generation sequence-variation data.

Scientific Applications:

• Differential diagnostics for rare diseases: The HPO-annotated corpus and models support phenotype-based differential diagnosis and comparison across rare disease annotations.
• Common disease modeling and annotation: Frequency-based HPO annotations across PubMed enable disease models for thousands of common diseases and analysis of phenotypic patterns.
• Translational and sequence-variation analysis: Phenotypic overlap analyses and HPO-driven annotations support phenotype-driven interpretation of next-generation sequence-variation data.
• Reference annotation resource: The platform aggregates over 250,000 phenotypic annotations covering more than 10,000 rare and common diseases for use in research and computational analyses.

Methodology:

Manual annotation of 228 abstracts with HPO concepts harmonized by three curators; construction of a standardized test suite of 32 test-case types covering 2,164 HPO concepts; frequency analysis of HPO disease annotations across >5 million PubMed abstracts with iterative optimization to improve precision and recall; evaluation of NCBO Annotator, OBO Annotator, and Bio-LarK CR against the annotated corpus and test suites.