KD4v

KD4v predicts the phenotypic impact of missense variants by generating interpretable rules that link variant annotations to deleterious or neutral outcomes.

Key Features:

• Phenotypic Effect Prediction: Classifies missense variants as deleterious or neutral based on learned rule-based models.
• Induction Logic Programming (ILP): Uses Induction Logic Programming (ILP) to learn symbolic, interpretable rules from annotated variant data.
• Interpretable Rules: Produces human-interpretable logical rules that explicitly relate variant attributes to phenotypic outcomes.
• Comprehensive Characterization: Integrates conservation data, physico-chemical properties, functional annotations, and 3D structural information for variant assessment.
• Annotated Variant Dataset: Leverages a dataset of missense variants annotated with multiple biological attributes to inform model learning.

Scientific Applications:

• Genomic Medicine: Aids identification and interpretation of potential disease-causing missense mutations for clinical and research investigation.
• Functional Genomics: Enables analysis of how specific amino-acid substitutions affect protein function and gene annotation.
• Evolutionary Biology: Uses conservation information to identify functionally constrained residues and infer evolutionary pressures on proteins.

Methodology:

KD4v trains Induction Logic Programming models on a dataset of missense variants annotated with conservation, physico-chemical properties, functional roles, and 3D structural information to learn rules that correlate these annotations with binary phenotypic outcomes (deleterious or neutral).