Extended Data Fig. 5: ESC-based BRCA2 SGE outperforms computational meta predictors.
From: Saturation genome editing-based clinical classification of BRCA2 variants
a, ROC curves indicate the performance of computational models at categorizing ClinVar-reported missense variants in the BRCA2 CTDB domain. ROC shows an AUC value of 0.96 in classifying ClinVar variants and AUC value of 0.96 in accurately categorizing nonsense and synonymous SNVs. We observed a moderate concordance to other computational predictors like CADD, BayesDel, REVEL, AlignGV-GD grade, PRIOR scores, EVE and AlphaMissense. b, Correlation between SGE-derived function scores and computational metrics in determining ClinVar-reported BRCA2 SNVs. The colour code represents the benign, likely benign, pathogenic, and likely pathogenic class reported in ClinVar.
