An intergenic GWAS signal and a candidate gene
A consortium performs a genome-wide association study of fasting triglyceride levels in 12,000 adults. The SNP rs987654 (G>A) has a minor allele frequency of 0.18 and is associated with 0.12 mmol/L higher triglycerides per A allele (95% CI 0.08-0.16; p=4×10^-14). The SNP lies in an intergenic region 45 kb from GENE-X and 120 kb from GENE-Y. In liver expression quantitative trait locus data, the A allele is associated with 1.4-fold higher GENE-X mRNA (p=2×10^-8) but no change in GENE-Y mRNA. GENE-X encodes a transcription factor, and a mouse knockout has 30% lower plasma triglycerides. The region around the SNP is predicted to be a liver enhancer and is conserved across mammals. A science writer claims that the SNP causes high triglycerides and that carriers should be treated differently from non-carriers.
1. Assess the evidence that GENE-X mediates the association between rs987654 and triglycerides. What is strong, what is missing, and what alternative explanation remains?
Read the worked answer
The GWAS association is strong statistically (p=4×10^-14) but only 0.12 mmol/L per allele, so it is a small population-level effect. The liver eQTL links the A allele to higher GENE-X mRNA in a relevant tissue, and the mouse knockout lowers triglycerides, giving directionally consistent functional support. However, the SNP is intergenic, 45 kb from GENE-X; association with GENE-X expression could reflect linkage disequilibrium with the true causal variant rather than rs987654 itself. The eQTL is also observational, and liver bulk RNA may mix cell types. The mouse knockout shows GENE-X can affect triglycerides but not that this SNP alters GENE-X. Missing evidence includes experimental perturbation of the SNP or enhancer, allele-specific expression, chromatin interaction data, and replication in other ancestries. An alternative is that the SNP tags a different causal variant affecting GENE-Y or another gene; the lack of a GENE-Y eQTL weakens but does not eliminate that possibility.
2. Explain why the SNP should not be described as a cause of high triglycerides in an individual. Quantify its expected contribution using the allele frequency and effect size.
Read the worked answer
A cause in the deterministic sense would mean carriers reliably develop high triglycerides. Here each A allele raises the mean by 0.12 mmol/L, with a 95% CI of 0.08-0.16, while triglyceride values vary widely with diet, adiposity, other genes, and measurement error. The allele frequency is 0.18, so genotype frequencies are approximately AA 0.032, AG 0.295, and GG 0.672. The additive variance contributed is 2pqβ² = 2(0.18)(0.82)(0.12)² ≈ 0.00425 mmol²/L². If the trait SD is about 0.5 mmol/L, that is 0.00425/0.25 = 1.7% of phenotypic variance. So the variant shifts the distribution slightly; it is not a diagnosis. Population attributable risk can be larger, but individual prediction needs a polygenic score and environment. The writer's claim confuses a statistical association with individual causation.
3. Propose two experiments that would test whether the A allele at rs987654 alters GENE-X expression through enhancer activity.
Read the worked answer
First, clone the 1-2 kb region containing each allele upstream of a minimal promoter and luciferase in a human hepatocyte line such as HepG2 or primary hepatocytes. If the A allele increases reporter activity relative to G, the region has allele-dependent enhancer activity. Include a promoterless control and an enhancer-deleted control, and test the region in a non-liver line to check tissue specificity. Second, use CRISPR-Cas9 to introduce the A allele into a G/G hepatocyte line, or delete the enhancer, and measure GENE-X mRNA by RT-qPCR and allele-specific expression by RNA-seq. A mouse knock-in of the human enhancer allele would test the effect in vivo. Chromosome conformation capture could confirm physical contact between the enhancer and the GENE-X promoter, and ATAC-seq could show chromatin accessibility. These experiments distinguish rs987654 from a linked bystander variant.