EmbedPVP

EmbedPVP prioritizes genomic variants from whole-exome and whole-genome sequencing by integrating neuro-symbolic deep learning with multi-source biological knowledge to relate variants to phenotypic consequences.

Key Features:

• Neuro-symbolic, knowledge-enhanced learning: EmbedPVP employs a neuro-symbolic framework that integrates deep learning with symbolic reasoning to leverage background knowledge about molecular mechanisms and phenotypic outcomes.
• Integration of genomic and clinical data: It combines genomic information with clinical phenotype data from human studies and model organisms for variant prioritization.
• Incorporation of diverse biological knowledge: The approach systematically incorporates gene functions, gene product roles, and anatomical sites of gene expression linked to phenotypic effects.
• Accounting for phenotype variability: EmbedPVP models variability in clinical phenotype assignment to improve robustness across diverse clinical presentations.

Scientific Applications:

• Genetic diagnostics for rare diseases: Prioritizes candidate variants to aid identification of disease-causing variants in undiagnosed patients using whole-exome and whole-genome sequencing data.
• Gene discovery and phenotypic interpretation: Supports identification of candidate genes and interpretation of complex phenotypes arising from multiple gene interactions using human and model organism data.

Methodology:

EmbedPVP applies neuro-symbolic learning that combines deep learning with symbolic reasoning using a large corpus of background knowledge—molecular mechanisms, gene functions, gene product roles, anatomical expression sites, and phenotypic consequences—from human studies and model organisms to relate genomic variants to phenotypic effects and prioritize variants.