pBRIT

pBRIT prioritizes candidate disease genes by integrating heterogeneous biological annotations and applying statistical models to map functional features to phenotypic annotations.

Key Features:

• Data Integration: Incorporates PubMed abstracts, Gene Ontology (GO), BLAST-based sequence similarities, Mouse Phenotype Ontology (MPO), Human Phenotype Ontology (HPO), pathway databases (Biocarta, KEGG, Reactome), and protein–protein interactions from PhosphoPOINT, CORUM, and IntAct.
• Information-Theoretic Modeling: Employs an Information-Theoretic approach to model feature dependencies and mitigate data sparsity and inter-feature biases.
• Bayesian Ridge Regression: Uses Bayesian Ridge regression to establish a linear mapping between functional and phenotype annotations based on the hypothesis that genes associated with similar diseases exhibit shared characteristics.
• Performance Evaluation: Evaluated against nine existing methods using over 2000 HPO–gene associations, achieving AUCs of 0.92–0.96 on benchmark datasets and an AUC of 0.80 on time-stamped HPO entries.
• Scalability and Stability: Designed for fast, scalable implementation and maintains stable performance despite changes in underlying annotation data.

Scientific Applications:

• Disease Gene Prioritization: Ranks candidate genes associated with human diseases using integrated annotations and statistical modeling with HPO-driven phenotypic concordance.
• Exome Interpretation: Prioritizes genes and variants from exome sequencing datasets to narrow candidate lists for follow-up.
• Translational Research and Target Discovery: Facilitates targeted genetic studies and aids identification of potential therapeutic targets by reducing candidate gene sets.

Methodology:

Integrates annotations from PubMed, GO, BLAST-based sequence similarities, MPO, HPO, Biocarta, KEGG, Reactome, PhosphoPOINT, CORUM, and IntAct; applies an Information-Theoretic approach to model feature dependencies and address data sparsity; uses Bayesian Ridge regression to learn a linear mapping between functional and phenotype annotations; and evaluates performance using AUC metrics on benchmark and time-stamped HPO gene associations.