Estimating the proportion of nonsense variants undergoing the newly described phenomenon of manufactured splice rescue

Abstract

Abstract A recent report described a nonsense variant simultaneously creating a donor splice site, resulting in a truncated but functional protein. To explore the generalizability of this unique mechanism, we analyzed > 115,000 nonsense variants from different databases using SpliceAI annotations. Between 0.61% (donor gain delta score > 0.8, for high precision) and 2.57% (> 0.2, for high sensitivity) of nonsense variants were predicted to create new donor splice sites at or upstream of the stop codon. These variants were less likely than all remaining nonsense variants in the same genes to be pathogenic in ClinVar (p < 0.001). Up to 1 in 40 nonsense variants may create new donor splice sites, potentially allowing for loss-of-function evasion through “manufactured splice rescue”. We urge caution when interpreting nonsense variants where manufactured splice rescue is a strong possibility and correlation with phenotype is challenging, as will often be the case with secondary findings and newborn genomic screening programs.