Orienting the causal relationship between imprecisely measured traits using GWAS summary data.

Area of Science:

• Genetics
• Biostatistics
• Molecular Biology

Background:

• Causal inference tests (CIT) combine genetic associations and mediation for trait causality.
• Measurement error in phenotypes can cause CIT to infer incorrect causal directions, amplified by larger sample sizes.
• This limitation may extend to other mediation-based causal inference approaches.

Purpose of the Study:

• To introduce an extension of Mendelian randomization (MR) for robust causal direction inference between traits.
• To address limitations of existing methods, particularly susceptibility to measurement error and confounding.
• To apply the novel method to infer causality between DNA methylation and gene expression.

Main Methods:

• Developed an extension to Mendelian randomization (MR) using summary-level data from genome-wide association studies.
• The method is designed to be less susceptible to bias from measurement error and unmeasured confounding.
• Applied the MR extension to investigate the causal relationship between DNA methylation and gene expression.

Main Results:

• The novel MR extension can infer causal direction using only summary-level GWAS data.
• The method demonstrates reduced susceptibility to bias from measurement error and confounding.
• Inferred causal direction between DNA methylation and gene expression, with DNA methylation often being causal.
• Results are sensitive to platform-specific measurement error differences and horizontal pleiotropy.

Conclusions:

• The extended MR approach offers advantages for causal inference, especially with summary-level data.
• While generally suggesting DNA methylation as causal for gene expression, biases must be considered.
• Triangulating results from MR and other methods like CIT with sensitivity analyses is crucial for reliable causal conclusions.