Gene-specific machine learning model EpiPred identifies likely pathogenic variants in the epilepsy-related gene STXBP1.

STXBP1

Machine-learning classifier calibrated on STXBP1 alone, with predictions tested against cellular assays of protein abundance, solubility, stability and interaction with syntaxin 1

STXBP1 variants are a frequent cause of early-onset developmental and epileptic encephalopathies and related neurodevelopmental disorders, yet most reported missense variants remain classified as variants of uncertain significance. EpiPred is a gene-specific machine-learning classifier, trained on a curated set of pathogenic and benign variants, that predicts the pathogenicity of STXBP1 missense variants. The authors report that it outperformed global prediction tools in accuracy, sensitivity and specificity, and they tested its predictions against assays measuring protein abundance, solubility, stability and interaction with the SNARE partner syntaxin 1. These biochemical readouts aligned closely with model outputs and led to proposed reclassification of several possibly misdiagnosed variants, which warrant further validation and clinical reevaluation. The model is deployed in an interactive web application covering every possible STXBP1 missense variant.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

Calibrating a predictor gene by gene and then testing it against orthogonal functional assays is the right answer to the VUS problem — and also the most expensive one, since training set and assays must be rebuilt for every gene. The abstract reports no calibration of the score into ACMG/AMP evidence strength: until a PP3/BP4 strength is established, the output informs discussion without entering the classification workflow directly. The immediate stake is eligibility for protein stabilizer trials, which requires precisely a classified variant.

Analysis by Dr Thibaut Benquey