DeepPVP

DeepPVP prioritizes likely causative variants in whole exome and whole genome sequence datasets by integrating phenotype-based semantic similarity with pathogenicity predictions from deep neural networks.

Key Features:

• Automated inference and deep learning: Integrates automated inference with deep neural network architectures to generate pathogenicity predictions.
• Phenotype-based prioritization: Combines pathogenicity prediction with semantic similarity measures to assess phenotypic similarity between patient and disease-associated phenotypes.
• Input scope: Operates on whole exome (WES) and whole genome (WGS) sequence datasets to rank candidate variants.
• Identification of causative variants: Prioritizes variants that are likely involved in the pathogenesis of a patient’s phenotype.
• Performance superiority: Empirical evaluations report improved speed and accuracy compared to existing phenotype-based variant prioritization methods.

Scientific Applications:

• Clinical genomics: Prioritizing pathogenic variants to inform genetic diagnosis and clinical decision-making.
• Personalized medicine and research: Investigating genetic underpinnings of complex diseases and supporting the development of individualized interventions.

Methodology:

Measures semantic similarity between patient phenotypes and known disease-associated phenotypes and integrates these measures with pathogenicity predictions produced by deep neural networks.