PSC readme

Description of analysis:
We conducted GWAS on low-dimensional representations (LDRs) of voxel-level fractional anisotropy (FA) data for UKB subject (Phases 1-6), where FA was derived from the PSC pipeline. In total, there were 6,090 fibers consisting of 100 voxels each (609,000 total voxels). These were clustered into 430 fiber clusters prior to conducting GWAS. For each cluster, we constructed LDRs on the Phase 1-5 cohort (discovery) and Phase 6 cohort (validation), where the functional principal component analysis (fPCA) bases were derived from the discovery cohort. Voxel-level GWAS reconstruction requires (1) LDR summary statistics, (2) fPCA bases, and (3) estimated LDR variance-covariance matrices. The voxel-level GWAS results included here post-screened based on both statistical significance (correcting for total effective number) and spatial significance (null distribution of cluster size, i.e. number of significant associations in given fiber cluster). For the validation voxel-level GWAS, we only included subset of candidate SNPs identified in discovery.

Total effective number: 5479.48

File structure:
Fiber_atlas_100points_FullName.csv: File with the assigned fiber cluster for each fiber. Assignment was based on the \’93Clust_Dis30mm\’94 column. Fibers within a cluster share the same network pair

nLDR.csv: File containing selected number of LDRs for each fiber cluster (CID), for both the discovery cohort (nLDR_disc) and the validation cohort (nLDR_val). The number of LDRs were selected to preserve image variation (80-90%) and ensure high correlation between raw and reconstructed images (0.85-0.95)

eff_num.csv: File containing the effective numbers per fiber cluster (CID), plus the quantiles of cluster sizes used for screening voxel-level GWAS results (quantile_disc and quantile_val for discovery and validation, respectively)

coords/: Folder containing txt files with coordinates for each voxel within a given fiber cluster. File names have convention coords/coords_fiberClust_X.txt, where X is the fiber cluster from 1-430

LDR/FA/: Folder containing LDR GWAS results for discovery (phase1to5) and validation (phase6) cohorts
phase1to5/fpca/: Contains fPCA bases derived for discovery cohort. File names UKB_FA_phase15_X_bases_top, where X is fiber cluster from 1-430 phase1to5/ldr/: Contains LDR covariance matrices for discovery cohort. File names UKB_FA_phase15_X_ldr_cov_top, X=1-430
phase1to5/sumstats/: Contains LDR summary statistics for discovery cohort, including .sumstats files with effects estimates, and .snpinfo with variant information. File names UKB_FA_phase15_X*, X=1-430

phase6/ldr/: Contains LDR covariance matrices for validation cohort. File names UKB_FA_phase6_X_ldr_cov_top*, X=1-430
phase6/sumstats/: Contains LDR summary statistics for validation cohort, including .sumstats files with effects estimates, and .snpinfo with variant information. File names UKB_FA_phase6_X*, X=1-430
Note: No phase6/fpca/ sub-folder since we use the Phase 1-5 bases for both discovery and validation

Voxel/FA/: Folder contains post-screened voxel-level GWAS results for discovery (phase1to5) and validation (phase6) cohorts
phase1to5: Discovery cohort voxel-level GWAS results. File names UKB_FA_phase15_X_sig_vgwas_screen.txt, X=1-430* (13 fiber clusters missing, see below)
phase6: Replication cohort voxel-level GWAS results. File names UKB_FA_phase6_X_sig_vgwas_screen.txt, X=1-430* (13 fiber clusters missing, see below)
Note: Voxel-level GWAS was post-screened as described in “Description of analysis”. For reproducing results, refer to tutorial at https://docs.google.com/document/d/1oXQkdN-oTD6RcY29ukQ8dZAq0nENB54Hrm-8ny_lhik/edit?tab=t.0
Note: 13/430 fiber clusters, no significant associations remained after post-screening. These fiber clusters are not contained in the Voxel/FA/ results files.