GABAA receptor functional variant effect prediction using multi-task phenotypic learning (GENTLY): Precomputed scores

Precomputed scores for all possible missense variants in genes encoding GABAA receptor subunits. The set of all possible variants is derived from AlphaMissense and based on MANE Select transcripts, with reference IDs provided in the source publication. Columns: gene, aa1, pos, aa2: Variant identifier; prediction: binary prediction of loss-of-function (LOF) or gain-of-function (GOF); pGOF, pLOF: Class probabilities from Platt scaling; distance: Distance from decision boundary (SVM hyperplane), a sigmoid distribution used to derive class probabilities; top_percentile: boolean, is this variant in the top score percentile for pGOF/pLOF? neutral_range: boolean, is this variant in the score range commonly observed for population variants? ACMG_evidence_pLOF: calibrated evidence strength for the PP3 criterion These scores are from the structural (non-clinical) model and should be used under the general framework proposed in the source publication. This score and model are for research use only and should not be used in diagnostics or for clinical decision-making. For all further information, please see the source publication.