Questions: Genetic Epidemiology: Heritability and Gene-Environment Interaction

Question 1

Average height in the Netherlands increased by roughly 20 cm over the 20th century, yet height has an estimated heritability of ~0.8. Which interpretation is correct?

• A. The heritability estimate must be wrong, since genetic traits cannot change this fast
• B. The increase proves that height is mostly environmental, contradicting the heritability estimate
• C. There is no contradiction: heritability measures variance within a population in a given environment, and improved nutrition altered the environment for everyone
• D. Heritability of 0.8 means 80% of any individual's height is genetically determined, so the environmental change only explains 20% of the increase

Answer: (C)

Heritability is a population-level statistic measuring what fraction of trait *variance* in a given population and environment is attributable to genetic variance. A high heritability says nothing about whether the trait can change when the environment changes — it only says genes explain much of the *differences between people* in the current environment. When nutrition improved uniformly, it shifted the entire distribution upward without touching heritability, which depends on relative differences remaining genetically explained.

Question 2

A GWAS for Type 2 diabetes identifies 100 SNPs, each with an odds ratio around 1.10, that together explain only 15% of the estimated heritability (~50%). What is the most likely explanation?

• A. The heritability estimate from twin studies must be inflated by shared environments, so the true heritability is ~15%
• B. Missing heritability likely comes from rare variants, gene-gene interactions, and limitations of additive variance decomposition not captured by common SNP arrays
• C. The GWAS significance threshold of p < 5×10⁻⁸ is too stringent and is excluding real variants
• D. GWAS cannot detect genetic contributions to complex diseases, so these 100 SNPs are likely false positives

Answer: (B)

The 'missing heritability' puzzle is a real and active research area. GWAS arrays are designed around common variants, so rare variants contributing to heritability are systematically missed. Gene-gene (epistatic) interactions and gene-environment interactions also contribute variance that standard additive heritability models don't fully capture. Option A is a genuine concern but doesn't explain the full gap; Option C is incorrect because the threshold is appropriate for genome-wide multiple testing correction.

Question 3

Heritability (h²) is a property of a population in a specific environment, not a fixed property of a gene or trait.

• T. True
• F. False

Answer: True

Heritability is defined as the proportion of population variance in a trait attributable to genetic variance, and both quantities depend on the environment. If everyone is exposed to the same uniform environment, environmental variance drops, and heritability appears to rise — not because genes became more important, but because the denominator (total variance) changed. The same trait can have different heritability estimates in different populations or environments.

Question 4

A high heritability estimate for a disease means that environmental interventions (diet, lifestyle, medications) will be ineffective because the disease is primarily genetic.

• T. True
• F. False

Answer: False

This is the most common misinterpretation of heritability. High heritability describes the sources of variance *between individuals in the current environment* — it says nothing about whether changing the environment could shift the entire distribution. PKU (phenylketonuria) is nearly 100% heritable yet completely preventable by removing phenylalanine from the diet. Heritability quantifies explanation of existing variation, not immutability.

Question 5

Why does studying gene-environment (G×E) interaction require larger sample sizes than studying either genetic or environmental main effects alone?

Model answer: Interaction effects — where the influence of a genetic variant depends on the presence of an environmental exposure — are smaller in effect size and harder to detect than main effects. Statistical tests for interactions have lower power because you are looking for differences in differences: does the genetic variant increase risk more in exposed than unexposed individuals? This requires stratifying the sample by exposure status, which reduces the effective sample size for each stratum, and the interaction term adds an extra degree of freedom.

In practical terms, a GWAS with 100,000 participants may have sufficient power to detect a main-effect SNP with OR = 1.10, but detecting a G×E interaction where the OR is 1.15 in exposed and 1.00 in unexposed might require 500,000+ participants. This is why G×E studies have historically been underpowered and results poorly replicated — the phenomena are real but demand biobank-scale data.