teaguesterling/plinking_duck v0.2.0

v0.2.0 — GWAS regression and filter pushdown

on GitHub

What's New

GWAS Regression (plink_glm)

• Per-variant association testing using plink2's regression engine
• Linear regression for continuous traits, logistic regression for case/control
• Automatic Firth correction when logistic regression fails to converge
• Covariate support (age, sex, PCs, etc.)
• Phenotype and covariate loading from .psam columns or external lists

Filter Pushdown

• af_range — filter variants by allele frequency (uses PgrGetCounts, no decompression)
• ac_range — filter variants by allele count
• genotype_range — filter individual genotype values (non-matching set to NULL)
• All filters work across variant, genotype, and sample orient modes

Genotype Output Options

• orient := 'sample' — one row per sample with genotype array across variants
• genotypes := 'columns' — pivot mode with one column per sample/variant
• genotypes := 'list' — LIST(TINYINT) fallback for large cohorts
• phased := true — haplotype pair output as ARRAY(TINYINT, 2)
• dosages := true — dosage output support
• plinking_max_matrix_elements config option for sample orient memory guard

Performance

• Parallel per-sample accumulation for plink_missing (sample mode) and plink_score

Platform

• macOS amd64 + arm64 build fix (CMake -include flag deduplication)
• Windows MinGW CI disabled pending further investigation

Full Changelog

v0.1.2...v0.2.0