Refinement of the classification of DDX41 variants through analysis of aggregated clinical datasets

Deleterious germline DDX41 variants are the leading cause of heritable predisposition to myelodysplastic neoplasia and acute myeloid leukemia (MDS/AML). Accurate classification of pathogenicity is crucial for managing patients and their families. The absence of specific guidelines, along with late-onset disease, incomplete penetrance, and founder variants, poses challenges in clinical and laboratory practice. We aggregated a synthetic cohort (ASC) of DDX41 germline and somatic variants from 35 studies, including 1796 cases among 53686 patients, plus an additional 832 cases from non-cohort publications. We aimed to leverage the DDX41-ASC to develop and refine ACMG/AMP criteria on case enrichment (PS4), somatic associations (PP4), and computational prediction (PP3/BP4). Analysis confirmed that deleterious germline DDX41 variants are most common in MDS/AML. A quasi-case-control study with ancestry matching revealed overestimated odds ratios for variants in underrepresented groups. Exploiting germline-somatic associations, we developed a Bayesian multinomial model that updates the odds of pathogenicity based on the presence and number of somatic patterns. Comparison of prediction tools showed that AlphaMissense outperformed REVEL in sensitivity. These results were integrated into an online tool to facilitate the consistent application of criteria. Overall, this comprehensive analysis of DDX41-ASC provides an evidence framework to inform the development of DDX41-specific curation guidelines.