Causal graphs for the analysis of genetic cohort data

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Genome-wide association studies (GWAS) identify genetic associations with phenotypic traits.
• Association does not imply causation, necessitating methods to infer causality.
• Mendelian randomization (MR) is a popular method using GWAS summary statistics to infer causal effects.

Purpose of the Study:

• To review GWAS and MR from a causal inference perspective.
• To build intuition for causal diagrams in genetic research.
• To discuss assumptions, limitations, and potential biases in causal inference methods.

Main Methods:

• Graphical causal models to represent causal assumptions.
• Review of Mendelian randomization (MR) techniques.
• Examination of confounding by ancestry and selection bias.

Main Results:

• Causal graphs clarify assumptions and limitations of genetic association studies.
• Understanding causal assumptions is crucial for clinical relevance.
• Approaches for handling confounding by ancestry and selection bias are discussed.

Conclusions:

• Transparent communication of causal assumptions is vital for interpreting genetic findings.
• Causal inference methods, like MR, require careful consideration of underlying assumptions.
• Addressing confounding and selection bias enhances the reliability of causal conclusions from genetic data.