Causal inference and GWAS: Rubin, Pearl, and Mendelian randomization

Area of Science:

• Quantitative genetics
• Statistical genetics
• Causal inference

Background:

• Genome-Wide Association Studies (GWAS) are widely used for complex traits but interpreting results for biological processes is difficult.
• Variability in trait responses, species, breeds, and lack of follow-up studies hinder interpretation of GWAS.
• Appropriate statistical tests are essential to identify causal genes responsible for observed genetic variability.

Purpose of the Study:

• To review theoretical aspects of Rubin's Causal Model (RCM) and Pearl's Causal Inference (PCI).
• To discuss assumptions for randomization in RCM with observational data, focusing on Stable Unit Treatment Analysis (SUTVA).
• To introduce a novel statistical test for causal effects using independent predicted residual breeding values from animal models.

Main Methods:

• Review of Rubin's Causal Model (RCM) and Pearl's Causal Inference (PCI).
• Discussion of randomization assumptions for RCM with observational data and SUTVA.
• Development of a causal effects test using independent predicted residual breeding values from animal models.

Main Results:

• The proposed method using independent predicted residual breeding values from animal models overcomes confounder effects and population structure.
• This approach facilitates the identification of causal genes influencing genetic variability in complex traits.
• The study also examines whether additive genetic effects, as defined by Fisher and Falconer, can be considered causal under RCM or PCI.

Conclusions:

• Interpreting GWAS for complex traits requires robust causal inference methods.
• The proposed statistical test offers a powerful tool to identify causal genes by mitigating confounding factors.
• Further investigation into the causal nature of additive genetic effects is warranted.