Read this topic

If a disorder like schizophreniaA long-term mental illness involving episodes of psychosis such as hallucinations and delusions.Read more → runs in families but no single gene fully explains it, how do researchers even begin to look for the genes involved, without first guessing which one to check? A genome-wide association study (GWAS) is a study design that scans hundreds of thousands to millions of common genetic markers, called single nucleotide polymorphisms (SNPs, single-letter spelling differences in DNA that are common in the population), across the entire genome in large groups of people with and without a trait, testing each marker for a statistical association with that trait without first assuming which gene matters. For example, imagine researchers compare the DNA of 40,000 people with schizophrenia against 40,000 people without it, testing each of a million SNPs one at a time for a difference in frequency between the two groups. In practice, this means a GWAS can flag a genomic region as statistically linked to a disorder even when nobody had a prior theory about that region, which is exactly why it needs very large samples and a strict statistical bar before a finding is trusted.

A genome-wide association study genotypes a large panel of common SNPs (typically hundreds of thousands, with the rest of the genome's variation inferred through linkage disequilibrium, meaning nearby SNPs tend to be inherited together) in large case and control cohorts, then tests each SNP independently for a difference in allele frequency between groups. Because so many independent statistical tests are run, a conventional single test's significance threshold (p<0.05) would produce huge numbers of false positives by chance alone, so GWAS uses a much stricter, genome-wide significance threshold, conventionally p<5x10⁻⁸, which corrects for testing roughly a million independent common variants. Because common variants each carry only a small individual effect on risk, detecting them reliably requires very large sample sizes, often tens of thousands to hundreds of thousands of participants, which is why psychiatric GWAS findings only became robust once international consortia pooled cohorts. Results are conventionally displayed as a plot of statistical significance against chromosomal position, with clusters of significant SNPs appearing as peaks; because of linkage disequilibrium, a significant peak marks a genomic region, not necessarily the causal gene itself, so a significant SNP near a gene does not prove that gene is responsible. GWAS findings are combined across many variants into a polygenic risk score (PRS), a single number summing the weighted effect of many risk variants to estimate an individual's overall genetic liability, though PRS currently explains only a modest proportion of variance for most psychiatric disorders and is not used for individual clinical prediction. A GWAS signal on its own establishes only a statistical association across a population; confirming that a flagged gene actually causes the trait requires a separate functional-validation step, such as directly manipulating the candidate gene in a model organism.

A vignette describing researchers scanning very large numbers of genetic markers across the whole genome in thousands of cases and controls, with no single gene named in advance, looking for a statistical association with a psychiatric trait, often followed by mention of a strict significance threshold or the need for replication in an independent cohort.

• Candidate-gene study, a candidate-gene study tests one or a few genes chosen in advance on a prior biological hypothesis, unlike a GWAS, which scans the entire genome without assuming which gene is involved, at the cost of needing a far larger sample and a much stricter significance threshold.
• Linkage analysis, linkage analysis follows a rare, large-effect variant that co-segregates with a disorder within affected families, whereas a GWAS detects common, small-effect variants across large numbers of unrelated cases and controls.
• Whole-genome or whole-exome sequencing, sequencing reads the actual DNA letters directly and can detect rare or entirely novel variants, unlike a GWAS, which genotypes only a pre-selected panel of known common SNPs and infers the surrounding sequence through linkage disequilibrium.
• Polygenic risk score, a GWAS is the discovery study that identifies which variants are statistically associated with a trait, whereas a polygenic risk score is the downstream tool that sums the weighted effects of many of those variants to estimate an individual's overall genetic liability.

• Treating a significant GWAS hit as proof that the nearest gene causes the disorder — linkage disequilibrium means a significant signal marks a genomic region, not necessarily the causal gene, and confirming causation needs a separate functional-validation step (for example, editing the candidate gene in a model organism).
• Assuming a GWAS uses the same significance threshold as an ordinary hypothesis test (p<0.05) — because a GWAS tests roughly a million independent variants at once, it uses a far stricter genome-wide significance threshold (conventionally p<5x10⁻⁸) to keep the false-positive rate under control.
• Confusing a GWAS with a candidate-gene study because both look for gene-disease links — a GWAS is hypothesis-free and scans the whole genome, whereas a candidate-gene study tests one or a few genes chosen on a prior hypothesis and needs a far smaller sample, at the cost of being far more prone to non-replication.

When a stem describes a hypothesis-free scan of the whole genome for statistical association with a disorder, across large case-control samples, identify it as a GWAS rather than a candidate-gene study (which names a specific pre-chosen gene) or a linkage study (which follows inheritance within families). If the stem then asks what a significant GWAS finding proves, remember it only establishes statistical association with a genomic region; proving that a specific gene causes the trait requires further functional work, not the GWAS result alone.