Voxel-level summary statistics of hippocampus shape and white matter microstructure

This deposit hosts voxel-level GWAS summary statistics of hippocampus shape and white matter microstructure using 33,000 UKB unrelated white subjects. The data was generated by using the highly-efficient imaging genetics (HEIG v1.0.0) framework where only the triplets - summary statistics of low-dimensional representations (LDRs), the functional bases, and the inner product of covariate-effect-removed LDRs - are shared, which is sufficient to recover all voxel-variant pairs as well as to conduct voxel-level heritability and (cross-trait) genetic correlation analysis. Check the tutorial and the example data used in the tutorial. Check code for how to use the data to reproduce the results in the manuscript (not available yet). The shared data includes: 1. Triplets for hippocampus shape measured by the radial distance from the medial model for each vertex. The original images contain 15,000 vertices. Left and right hemispheres were analyzed separately. 2. Triplets for 21 white matter tracts measured by fractional anisotropy. The original images contain 88 ~ 3503 voxels. Each tracts were analyzed separately. 3. LD matrix and its inverse for 22 chromosomes including 460k genotyped SNPs. LD matrix and its inverse were estimated by using two separate datasets each containing 8.4k white unrelated subjects in UKB. Two regularization levels are provided: {85%, 80%} for heritability and genetic correlations within images and {75%, 70%} for cross-trait genetic correlations. 4. LD matrix and its inverse for 22 chromosomes including 1.2 million imputed HapMap3 SNPs. LD matrix and its inverse were estimated by using two separate datasets each containing 42k white unrelated subjects in UKB. Two regularization levels are provided: {98%, 95%} for heritability and genetic correlations within images and {90%, 85%} for cross-trait genetic correlations.