Shape GWAS readme

Description:
We ran GWAS analysis using low dimensional Representations (LDR) on the image data. Imaging data was segmented eight structures:
accumbens, amygdala, caudate, hippocampus, pallidum, putamen, thalamus, ventricles for left and right hemispheres. The structures for
both sides contained 15000 vertices each. The only exception was the ventricles which contained around 23000 each. Visualizations available at here

Each structure was then measured by six features (f1-f2, f4-f7).
f1 – radial distance, f2 – mTBM1 (zoom level), f4 – mTBM3 (tilt/bend), f5 – determinant (surface area),
f6 – eigenvalue 1 (max), f7 – eigenvalue 2 (min)

f3 is excluded due to redundance

Measurements were done through Low Dimensional Representations(LDRs) then recovered at voxel level for phase1 through phase 5 data.
Phase 6 data was used for replication purposes. These LDR’s are created through functional PCA to contain 87-90% of the variance.

file structure:

The data is stored at LDR level by hemisphere, subcortical region, and feature
Number of LDR’s are stored in the .txt file.

Each is separated by sumstat and npy zip

sumstat.zip: contains a .sumstat and a .snpinfo files to reconstruct LDR level Summary statitics

npy.zip: contains coviates and bases in order to reconstruct voxel-level data using the HEIG pipeline

see more details in https://github.com/Zhiwen-Owen-Jiang/heig/wiki/5.3-Voxel%E2%80%90level-GWAS-reconstruction