HPO2GO

HPO2GO predicts associations between Human Phenotype Ontology (HPO) terms and Gene Ontology (GO) terms to link phenotypic abnormalities to underlying biomolecular functions.

Key Features:

• Cross-Ontology Mapping: Automates mapping between HPO terms (phenotypic abnormalities) and GO terms (biomolecular functions), where each association suggests a phenotype may arise from loss of a corresponding function.
• Data-Driven Approach: Leverages curated HPO and GO annotations and calculates frequency of co-annotations on the same genes or proteins to derive mappings.
• Statistical Filtering: Employs statistical resampling techniques to filter out mappings that may occur by chance.
• Biological Relevance Validation: Validates selected mappings through literature discussion to support biological significance.
• Predictive Capability: Predicts novel gene/protein–ontology term–disease relationships, including HPO term–protein associations (HPO2protein).
• Benchmarking Performance: Assesses predictive performance using the CAFA2 HPO prediction target protein set, reporting a maximum F-score of 0.35.
• Extensibility: Applies an automated cross-ontology mapping methodology that can be adapted to other ontologies.

Scientific Applications:

• Disease mechanism elucidation: Linking gene product functions to phenotypic abnormalities to study genetic underpinnings of disease.
• Gene/protein–phenotype association prediction: Generating candidate HPO term–protein associations for downstream experimental or computational follow-up.
• Ontology-based benchmarking: Evaluating ontology-driven prediction methods against established benchmarks such as CAFA2.

Methodology:

Uses curated HPO and GO annotation datasets; extracts mappings based on co-annotation frequencies on the same genes or proteins; filters mappings via statistical resampling; validates selected mappings via literature review; predicts HPO term–protein associations and evaluates performance against the CAFA2 benchmark.