The authors tested the burden of rare loss-of-function and deleterious missense variants gene by gene across biobank and disease-focused sequencing cohorts, using calibrated proxy phenotypes to gain power for late-onset disease. At study-wide significance they recovered eight established dementia genes and two Parkinson's genes, and reported one new dementia association and two new Parkinson's associations. The strongest new signal, ANKRD27, showed damaging variants clustering in the domains that mediate Rab GTPase and retromer interactions; the sample size is not stated in the abstract.
Why it is interesting: Rare-variant architecture bears on who develops dementia and on which pathways can be acted on, though an unreplicated preprint signal moves little on its own.