Bias correction for inverse variance weighting Mendelian randomization.

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Inverse-variance weighted two-sample Mendelian randomization (IVW-MR) is a primary method for causal inference using genome-wide association studies (GWAS) summary statistics.
• IVW-MR estimates can be biased by weak instruments, winner's curse, and sample overlap between exposure and outcome GWAS.
• Understanding and mitigating these biases is crucial for accurate causal effect estimation in genetic research.

Purpose of the Study:

• To develop a novel method (MRlap) that simultaneously corrects for weak instrument bias, winner's curse, and sample overlap in IVW-MR.
• To analytically derive and estimate bias corrections using only summary statistics.
• To improve the accuracy and reliability of causal inference in genetic association studies.

Main Methods:

• Developed MRlap, a method leveraging spike-and-slab genomic architecture and linkage disequilibrium score regression.
• Analytically derived bias correction factors for IVW-MR.
• Validated the method using extensive simulations across various realistic scenarios.
• Applied MRlap to obesity-related exposures with both nonoverlapping and fully overlapping samples.

Main Results:

• MRlap significantly reduced bias in simulated IVW-MR estimates, with reductions up to 30-fold in some scenarios.
• Bias correction effectiveness increased with larger sample sizes.
• Traits with low heritability and/or high polygenicity were found to be more susceptible to bias.
• Statistically significant differences were observed between IVW-based and MRlap-corrected causal effects for obesity-related exposures.

Conclusions:

• MRlap provides a robust analytical solution to correct for key biases in IVW-MR, enhancing causal inference accuracy.
• The method facilitates the use of larger GWAS datasets by accommodating sample overlap, thereby increasing statistical power.
• MRlap represents a significant advancement for robust causal effect estimation in genetic epidemiology.