A genome-wide association study (GWAS) tests thousands of loci for statistical association with a trait. Each pillar's height is -log10(p-value) for that locus; taller pillars are more significant. Because so many loci are tested at once, a strict genome-wide significance threshold (traditionally p < 5×10⁻⁸, a Bonferroni correction for ~1 million independent tests) is required to control false positives — the glowing threshold plane marks that cutoff.
Power ≈ Φ(√(2N·MAF(1-MAF))·β − z_(1-α/2))
Bonferroni threshold: α_corrected = α / (# loci tested)
-log10(p) increases with cohort size N, effect size β, and allele frequency
- Cohort size — number of simulated participants; larger cohorts shrink standard error and raise -log10(p) for real signals.
- True effect size — the underlying (simulated) strength of one locus's real association with the trait, driving one deliberately "true" peak.
- Risk allele frequency — how common the variant is in the population; rarer alleles need larger cohorts to detect at the same effect size.
- Significance threshold — moves the glowing cutoff plane; stricter thresholds (5×10⁻⁹) demand taller peaks to count as "significant."
- Resample cohort — draws a fresh random cohort, regenerating noise peaks and re-testing the true signal locus.
This statistical framework underlies real discoveries like the many GWAS-identified loci for height, type-2 diabetes risk, and Alzheimer's susceptibility.