Quantitative prior probabilities for disease-causing variants reveal the top genetic contributors in inborn errors of immunity

We present a novel framework for quantifying the prior probability of observing disease-associated variants in any gene for a given phenotype. By integrating large-scale genomic annotations, including population allele frequencies and ClinVar variant classifications, with Hardy-Weinberg-based calculations, our method estimates per-variant observation probabilities under autosomal dominant (AD), autosomal recessive (AR), and X-linked modes of inheritance. Applied to 557 genes implicated in primary immunodeficiency and inflammatory disease, our approach generated 54,814 variant probabilities. First, these detailed, pre-calculated results provide robust priors for any gene-disease combination. Second, a score positive total metric summarises the aggregate pathogenic burden, serving as an indicator of the likelihood of observing a patient with the disease and reflecting genetic constraint. Validation in NFKB1 (AD) and CFTR (AR) disorders confirmed close concordance between predicted and observed case counts. The resulting datasets, available in both machine-readable and human-friendly formats, support Bayesian variant interpretation and clinical decision-making. ¹