Lesson 2 GWAS Study Design and Traits

2.1 Why study design matters

In GWAS, study design decisions shape everything that follows.
Poor design cannot be fixed later with better models or stricter thresholds.

Before any genotypes are analyzed, you must be clear about:

GWAS works best when these questions are addressed explicitly and early.

Traits analyzed in GWAS generally fall into two broad categories.

Binary traits describe the presence or absence of a condition.

Examples include:

Binary traits are commonly analyzed using logistic regression models.

Quantitative traits take continuous values.

Examples include:

These traits are typically analyzed using linear regression models.

A phenotype is not just a column in a table.
It is an operational definition of the biological or clinical concept you want to study.

Good phenotype definitions are: - precise - reproducible - consistent across samples

Ambiguous or noisy phenotypes reduce power and increase false discoveries.

In case control designs, individuals are grouped based on disease status.

Key considerations include:

Imbalances between cases and controls can introduce bias that appears as genetic association.

For quantitative traits, attention should be paid to:

Trait distributions that strongly deviate from model assumptions can distort association results.

Covariates are variables included in the model to control for non-genetic effects.

Common covariates include:

Failing to include relevant covariates can lead to spurious associations. Including unnecessary covariates can reduce power.

GWAS typically detects variants with small effect sizes.
Adequate sample size is therefore critical.

Power depends on:

Underpowered studies are prone to false negatives and unstable results.