LaBranchoR predicted branchpoints overlap pathogenic variants and exclude common variants. (A) Overlaps with pathogenic variants for Mercer et al. (2015) high confidence branchpoints and LaBranchoR predictions for 3′SS with a Mercer et al. (2015) high confidence site. (B) Similar for Taggart et al. (2017) branchpoints. (C) Genome-wide pathogenic variant overlaps for the current state-of-the art branchpoint predictor, branchpointer (Signal et al. 2018), LaBranchoR, and LaBranchoR tuned to predict the same number of branchpoints as branchpointer (LaBranchoR +). (D) Comparison of the common variant frequency in ExAC for distance from 3′SS matched UNA trinucleotides, where the A is implicated as a branchpoint (BP) and not a branchpoint (not BP). (E) Enrichments in variation frequency of branchpoint UNAs, as compared to non-branchpoint UNAs for common and rare variants. (F) Similar comparison of branchpoints in high probability loss of function intolerant (pLI) genes to low pLI genes. (G) In silico mutagenesis (ISM) scores are defined as the change in score of our predicted branchpoint induced by the variant. (H) LaBranchoR ISM scores effectively classify pathogenic variants. A receiver–operator curve for HGMD and ClinVar variants sorted by LaBranchoR ISM scores and SPIDEX scores.
