VIPPID

VIPPID predicts the pathogenicity of single nucleotide variants (SNVs) in primary immunodeficiency diseases (PIDs) using gene-specific models to improve discrimination between pathogenic and non-pathogenic variants.

Key Features:

• Gene-specific predictive models: Implements distinct predictive models tailored to each of the most prevalent PID-associated genes.
• Classifier: Uses a Conditional Inference Forest model as the classification algorithm.
• Feature integration: Integrates 85 SNV features and scores derived from 20 existing prediction tools as model inputs.
• Genetic-context awareness: Captures unique characteristics and functional implications of mutations within specific genetic contexts.
• Algorithmic framework: Utilizes a novel algorithmic approach to examine how mutation features relate to encoded protein characteristics and functions.
• Performance: Demonstrated elevated predictive performance with an area under the curve (AUC) of 0.91 compared with non-specific prediction tools.

Scientific Applications:

• Clinical variant interpretation: Supports classification of SNVs in clinical genetic testing for primary immunodeficiency diseases by distinguishing pathogenic from non-pathogenic variants.
• Protein function and variant effect research: Enables investigation of mutation features to explore characteristics and functions of encoded proteins.
• PID-focused genetic research: Provides tailored predictive models applicable to research on the genetics of primary immunodeficiency diseases.

Methodology:

Builds distinct gene-specific models using a Conditional Inference Forest classifier that integrates 85 SNV features and scores from 20 existing prediction tools.