Command Line Interface

June 6, 2026 · View on GitHub

GraphLD provides a graphld CLI for LDGM-backed workflows and a separate estest CLI for enrichment score tests, which operate on precomputed scores instead of LDGMs. For the enrichment score test CLI, see Enrichment Score Test.

Subcommands

Shared Usage

Inspect the command surface with:

uv run graphld -h
uv run graphld <subcommand> -h

Most analysis subcommands follow this shape:

uv run graphld <subcommand> SUMSTATS OUT [options]

simulate is the exception: it takes an output summary-statistics path rather than an input summary-statistics file.

Shared Inputs

The main graphld analysis commands accept summary statistics in:

GraphLD uses LDGM metadata to locate LD blocks. By default this is:

data/ldgms/metadata.csv

Override it with --metadata when needed.

For annotation-driven workflows, GraphLD supports:

LDSC .annot files
UCSC .bed region annotations
GMT gene-set annotations via --gene-annot-dir

See File Formats for details.

Shared Options

Many subcommands share these options:

Option	Description
`-p, --population`	Restrict LDGM selection to a population such as `EUR` or `EAS`
`-c, --chromosome`	Restrict analysis to one chromosome
`-n, --num-samples`	Override sample size when it cannot be inferred from the input
`--num-processes`	Set the worker-process count
`--run-in-serial`	Disable multiprocessing
`--metadata`	Use a non-default LDGM metadata CSV
`-v, --verbose`	Print detailed progress
`-q, --quiet`	Suppress normal progress output

surrogates is population-specific. The CLI defaults to EUR; pass --population explicitly for other reference populations.

Shared Output Pattern

Commands that consume summary statistics typically write a single tab-separated output file or HDF5 artifact to the OUT path you pass on the command line.

The main exception is reml, which writes multiple result files with a shared prefix.