Back

A Bayesian model for quantifying genomic variant evidence sufficiency in Mendelian disease

Lawless, D.; The quantitative omic epidemiology group,

2025-12-04 genetic and genomic medicine
10.64898/2025.12.02.25341503 medRxiv
Show abstract

SummaryClinical genomic interpretation depends on heterogeneous evidence checks that vary across institutions and pipelines. This variation prevents evidence from being compared or verified in a reliable way. We address this logistical barrier by introducing a universal, verifiable layer that evaluates the completeness of available scientific evidence independently of upstream analytic systems. Quantitative Evidence Sufficiency (Quant ES) summarises, for each variant, how much verifiable evidence is available using a binary matrix derived from registered rule sets. A closed form Beta Binomial model produces a single evidence sufficiency estimate with a credible interval and genome wide percentile. This separates evidence availability from provider-specific logic and enables consistent interpretation and reuse of results across institutions. ImplementationQuantBayes implements the method as compiled C binaries and an R package. It requires no access to proprietary algorithms and scales to millions of variants, allowing integration into existing diagnostic and research workflows without modification to upstream systems. AvailabilityThe QuantBayes software releases for macOS and Linux are at https://doi.org/10.5281/zenodo.17919369. The QuantBayes R package is on the Comprehensive R Archive Network (CRAN) https://doi.org/10.32614/CRAN.package.quantbayes. All releases are under the MIT licence. Graphical abstractGenomic analyses produce candidate variants, but the completeness of their supporting evidence is not directly comparable across institutions. Quant Evidence Sufficiency operates on a binary evidence matrix to summarise how much verifiable scientific evidence is available for each variant. This separates analysis outputs from interpretation and delivers an interpretable, uncertainty-calibrated measure of credibility that can be shared and reused without exposing provider-specific methods or intellectual property. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=73 SRC="FIGDIR/small/25341503v3_ufig1.gif" ALT="Figure 1"> View larger version (20K): org.highwire.dtl.DTLVardef@2a7640org.highwire.dtl.DTLVardef@1b9808eorg.highwire.dtl.DTLVardef@1a721e2org.highwire.dtl.DTLVardef@1a1ad7d_HPS_FORMAT_FIGEXP M_FIG C_FIG

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.