Data

Data

This directory contains descriptions for the three datasets used. Please note this section exists solely to describe the source, format, and preprocessing steps performed on each cohort.

1. UK Biobank

• Source: Preprocessed and imputed data on teralab (/home/sharedFolder/referenceData/ukb/)
• Format: Binary PLINK files (.bed/.bim/.fam) and PCA projections
• Used as a benchmarking dataset to compare PRSice-2, PRScs, and pgsc_calc

2. Lyday

• Source: Raw VCF files from a collaborator.
• Processing Steps:
  • Chromosome-level VCF merging
  • Genotype harmonization using conform-gt
  • Imputation using Beagle 5.5
  • Conversion to PLINK format
• Notes: Output from these steps was used for PRS calculation with pgsc_calc.

3. All of Us

• Source: All of Us genomic dataset

• Processing Steps:

  • Building Case-Control Cohort using the All of Us cohort builder
  • Subsetting individuals by self-identified race/ethnicity (SIRE): EUR, AFR, AMR
  • Individuals with age & sex present in the reported data
  • Obtaining the scoring file containing variant IDs, variant position, ref/alt allele and effect size information
  • Filtering variants to create the variant table
    • Done using Cromwell workflow to downsample to ACAF variant data
    • Participant ID and the scoring file variants are extracted
  • Extracting PCs for ancestry normalization

All of Us Cohort distribution

All analysis was conducted on a remote server; data was not transferred locally or externally.
Please refer to the Tools/, Ancestry adjustment with pgsc_calc./, and Validation/ directories for additional implementation details.