Fig. 5: AbExp prioritizes deleterious variants.

a Precision-recall curve of AbExp (trained with and without using the biological coefficient of variation), CADD, and LOFTEE on distinguishing pathogenic from benign and likely begin variants in ClinVar. LOFTEE as a binary predictor is shown as a single point. The dashed vertical bars denote the high and low confidence cutoffs of AbExp. b Odds-ratio of high-impact variants among 4,921,131,336 absent, 62,134,299 singleton, 38,474,061 rare (MAF < 0.1%), and 14,438,258 common SNVs in gnomAD for the models shown in (a). The analysis is restricted to variants within 5 kb of protein-coding genes. Error bars show Wald 95% confidence intervals from logistic regression fits. The high-impact cutoffs for CADD and AbExp without BCV were set to match the quantile of the high-impact cutoff of AbExp. c As in (b) among SNVs absent in gnomAD as a function of gene LOEUF decile²². Genes with a high LOEUF are more tolerant to loss of function. Error bars as in (b). d BCV versus LOEUF across all genes and tissues. The black line shows a running median between LOEUF and BCV highlighting the two genes from Fig. 2d. The autosomal recessive gene LTBP3 has a low LOEUF, denoting a low loss-of-function tolerance. In contrast, the olfactory gene OR2W3 has a high LOEUF, denoting a large loss-of-function tolerance.