Saturation genome editing of DDX3X clarifies pathogenicity of germline and somatic variation

There is an urgent clinical need to improve genetic variant interpretation through systematic and comprehensive functional characterisation. Loss-of-function of DDX3X is a leading cause of neurodevelopmental disorders (NDD) in girls. DDX3X is also a somatic cancer driver gene but of uncertain mechanism. We performed saturation genome editing (SGE) on DDX3X, testing in vitro the functional impact over time of 12,776 variants comprising all coding and splice-region single nucleotide variants, codon-deletions and NDD-associated variants. We identified 3,432 functionally abnormal variants, which fell in three distinct classes. We trained a machine learning classifier to specifically identify functionally abnormal variants of NDD-relevance. This classifier had at least 97% sensitivity and 99% specificity to detect variants pathogenic for NDD, substantially out-performing in silico predictors, and resolving up to 93% of variants of uncertain significance. We also confirmed that DDX3X predominantly acts as a tumour suppressor gene in cancer, and that the vast majority of driver mutations could be detected as functionally abnormal in our SGE data. Systematic maps of variant effects generated in experimentally tractable, but not overtly pathophysiologically-relevant cell types have the potential to transform interpretation of both germline and somatic disease-associated variation and catalyse mechanistic insights of therapeutic relevance.