ClinPred

ClinPred predicts the pathogenicity of human nonsynonymous single nucleotide variants (SNVs) to prioritize disease-relevant missense variants for genomic research and clinical interpretation.

Key Features:

• Machine Learning Algorithms: Two machine learning algorithms integrate existing pathogenicity scores to discriminate benign and disease-causing variants.
• Incorporation of gnomAD: Allele frequency data from the Genome Aggregation Database (gnomAD) are used as an input feature rather than as fixed cutoffs.
• Training on ClinVar: Training uses confidently annotated disease-causing variants from the ClinVar database.
• Predictive Accuracy: Demonstrates highest area under the curve (AUC) and improved specificity and sensitivity across test datasets.
• Robustness Across Conditions: Maintains consistent performance across disease types (e.g., cancer and rare diseases) and mechanisms (gain or loss of function).
• Pre-computed Exome Scores: Provides pre-computed scores for all possible human missense variants in the exome.

Scientific Applications:

• Clinical Variant Interpretation: Prioritizes nonsynonymous SNVs for use in clinical diagnostics and variant classification.
• Genetic Disease Research: Supports genomic research focused on understanding the genetic underpinnings of diseases.
• Molecular Mechanism Studies: Aids research into molecular mechanisms of various conditions by highlighting likely pathogenic missense variants.
• Therapeutic Target Identification: Assists in identifying potential targets for therapeutic intervention based on predicted pathogenicity.
• Personalized Medicine: Contributes to personalized medicine approaches by enabling more precise variant assessment.

Methodology:

Employs two machine learning algorithms that integrate existing pathogenicity scores; uses gnomAD allele frequency as an input feature; trains on confidently annotated disease-causing variants from ClinVar; and provides pre-computed exome-wide missense variant scores.