Background covariance adjustment distills shared genetic architecture across neurodevelopmental and neurodegenerative disorders

GWAS increasingly reveal shared genetic influences across neurodevelopmental, psychiatric, and neurodegenerative traits. However, cross-trait genetic covariance derived from GWAS summary statistics can be inflated by sample overlap and other structured background effects, obscuring higher-order genetic organization. We extend PathGPS, a recently developed statistical method that estimates an adjusted genetic covariance by subtracting a background covariance learned from weakly associated variants, and then extracts reproducible low-rank structure using rotation and bootstrap aggregation. When applying to 15 phenotypes related to neurodevelopmental and neurodegenerative disorders, the adjusted analysis yields four stable clusters with an interpretable topology. Adjusting for background covariance, which appears to be related to traumatic life experiences, sharpens the cluster boundaries and substantially shifts the clustering result for post-traumatic syndrome disorder. Simulations with controlled overlap and structured background covariance show that PathGPS has improved factor recovery relative to substantially shifts the clustering result for post-traumatic syndrome disorder.