Fig. 1: Predicting the likelihood of a loss-of-function mechanism based on the structural properties of pathogenic missense variants.
a Overview of the mLOF score framework. The missense LOF likelihood score mLOF is calculated from empirical distributions of the metrics EDC (spatial clustering) and ΔΔGrank (energetic impact) in LOF and non-LOF genes. This score is then used to update gene-level mechanism-specific priors (pDN/GOF/LOF) established in an earlier study18. The final posterior scores (postDN/GOF/LOF) represent adjusted estimates of the likelihood that a gene exhibits a specific molecular disease mechanism, given the structural properties of its pathogenic missense variants. b Receiver operating characteristic (ROC) curves and area under the curve (AUC) values of the mLOF score, the prior mechanism probability for the gene, and the posterior mechanism-specific scores across the binary class pairs used to construct the priors. The analysis is split into all genes, using all pathogenic missense variants, and a subset of single-phenotype genes, where only variants linked to the specific OMIM phenotypes are considered. N is the number of genes in each group. mLOF toptimal shows the optimal ROC threshold. Source data are provided as a Source Data file.
