GC Correction Tools

This repository provides an R-only utility to apply Genomic Control (GC) to GWAS or SAIGE step‑2 summary statistics.

• scripts/gc_adjust_general.R: R script using optparse + data.table to construct a chi‑square statistic from common columns (CHISQ, Tstat/var, BETA/SE, p.value.NA, p.value), estimate one lambda with optional MAC/MAF filter, optionally use a fixed prespecified lambda, and output GC‑adjusted results.
• data/test_data.csv: small synthetic dataset to try the tool.

Quick start

Run GC adjustment (R)

You need R with packages data.table and optparse.

Rscript scripts/gc_adjust_general.R \
  --input data/test_data.csv \
  --out out/test_gc_r

This will write out/test_gc_r.tsv.gz and out/test_gc_r.lambda.tsv.

Use a fixed lambda (bypass estimation):

Rscript scripts/gc_adjust_general.R \
  --input data/test_data.csv \
  --out out/test_gc_r_fixed \
  --lambda_fixed 1.12

Multithreading

Set threads for data.table with --cpus (defaults to parallel::detectCores()):

Rscript scripts/gc_adjust_general.R \
  --input data/test_data.csv \
  --out out/test_gc_r_threads \
  --cpus 8

Usage details

See docs/USAGE.md for CLI options and examples. docs/METHODS.md summarizes the chi‑square fallbacks and lambda estimation.

Notes

• Lambda is clamped to >= 1 unless you pass --allow_deflation.
• When using SAIGE binary outputs, the primary p.value is SPA-based; the scripts warn if they must derive chi‑square from that.
• Optional filtering for lambda estimation supports MAC or MAF. If both are provided, MAC is preferred.