genome-wide association study
/ G-W-A-S, 'gee-wass' /
Suppose you want to know which spots in the human genome influence a trait like height or the risk of diabetes, but you have no idea which genes are involved. One approach is to compare huge numbers of people — say everyone with the condition versus everyone without it — and scan their DNA for any letter that tends to come along with the trait. Wherever one version of a spot is more common in affected people, that spot is statistically associated with the trait. That sweeping, hypothesis-free comparison is a genome-wide association study.
A GWAS works by genotyping hundreds of thousands to millions of common single-letter variants (SNPs) scattered across the genome in tens of thousands of people, then asking, for each variant, whether one version is found more often in people with the trait than expected by chance. Because it tests so many variants at once, it demands a brutally strict statistical threshold (around p less than 5 x 10^-8) to avoid being swamped by false positives. The output is a 'Manhattan plot', a skyline of dots where towers rising above the threshold mark genomic regions linked to the trait.
GWAS has been enormously productive, finding thousands of robust associations and revealing that most common traits and diseases are polygenic — shaped by many variants of tiny individual effect rather than one gene. But its single most important caveat, repeated like a mantra in the field, is that association is not causation. A flagged SNP usually is not the cause itself; it merely sits near, and is inherited along with, the real causal variant, which still must be hunted down. GWAS hits also rarely explain a trait's full heritability, and findings from one population can fail to transfer to another, so a GWAS points toward biology rather than proving it.
A GWAS of 100,000 people might flag a SNP near a gene as associated with type 2 diabetes; this points researchers to a region worth investigating, but the SNP itself is usually just a marker riding along with the true causal variant nearby.
GWAS finds statistical associations across the genome — pointers to biology, not proof of cause.
Association is not causation. A GWAS hit is usually a marker near the real causal variant, not the cause itself, and findings often fail to transfer between populations.