BLUP (blup)

June 6, 2026 ยท View on GitHub

Use graphld blup to compute infinitesimal-model effect-size weights.

Basic Usage

uv run graphld blup \
    /path/to/sumstats.sumstats \
    output.tsv \
    --heritability 0.1 \
    --num-samples 100000

Required Option

OptionDescription
-H, --heritabilityHeritability for the analyzed variant scope, between 0 and 1

Inputs And Output

  • Input summary statistics: .sumstats, .vcf/.vcf.gz, or .parquet; see Summary Statistics
  • Output: a tab-separated file of BLUP weights at the path you pass as OUT

BLUP also uses the shared LDGM options documented in the CLI overview.

Notes

  • BLUP assumes an infinitesimal effect covariance matrix D; GraphLD sets trace(D) equal to --heritability across the matched LDGM effect indices in the analyzed scope. If you restrict BLUP with --chromosome or --population, pass the heritability for that restricted scope.
  • If --num-samples is omitted, GraphLD infers sample size from the input file.
  • Use --population, --chromosome, and --run-in-serial to limit the scope of a run or simplify debugging.