BayPR

BayPR estimates prevalence ratios of genetic variants between affected and unaffected populations using empirical Bayesian methods to predict variant pathogenicity from population data alone.

Key Features:

• Empirical Bayesian Methodology: BayPR employs an empirical Bayes framework to model variant prevalence ratios and infer probabilities of pathogenicity.
• Two-distribution assumption: The method assumes prevalence ratios form two distinct distributions corresponding to pathogenic and benign variants.
• R package implementation: BayPR is implemented as an R package.
• Agnostic to Variant Type: The approach applies across various DNA variant types without requiring variant-specific functional data.
• Population-data-only prediction: It predicts pathogenicity using population prevalence data alone, without additional functional or clinical information.
• High Predictive Accuracy (CFTR benchmarks): In CFTR testing, BayPR classified 300 of 313 curated variants correctly and achieved an AUC of 0.99 for functionally-confirmed missense CFTR variants versus ten commonly used algorithms.
• Broad applicability across genes and conditions: BayPR assigned high disease-causing probabilities to pathogenic/likely pathogenic variants and low probabilities to benign/likely benign variants across eight genes linked to seven Mendelian conditions.

Scientific Applications:

• Variant Pathogenicity Prediction: Provides probabilistic assessments of variant pathogenicity based solely on population prevalence data.
• Mendelian Disorder Research: Aids identification and classification of disease-causing variants across genes associated with Mendelian conditions.

Methodology:

BayPR models prevalence ratios between affected and unaffected populations under a two-distribution empirical Bayes framework and assigns posterior probabilities of pathogenicity to each variant; it is implemented as an R package.