Bayesian Mendelian randomization methods for index trait bias correction in subsequent trait genome-wide association...

Area of Science:

• Genetics
• Epidemiology
• Statistical genetics

Background:

• Index-trait bias, also known as index-event bias, arises in genetic studies when analysis is conditioned on an incident trait.
• This conditioning can introduce bias into genetic associations with subsequent traits, potentially leading to inaccurate findings.

Purpose of the Study:

• To propose and evaluate two Bayesian Mendelian Randomization (MR) methods, Bayesian weighted MR (BWMR) and MR-HORSE, for correcting index-trait bias in genome-wide association studies (GWAS).
• To compare the performance of these novel Bayesian MR methods against existing bias correction techniques through simulation studies.

Main Methods:

• Development and application of two Bayesian Mendelian Randomization (MR) methods: Bayesian weighted MR (BWMR) and MR-HORSE.
• A simulation study was conducted to compare BWMR and MR-HORSE against established methods like MR-IVW, MR-WMED, MR-Egger, and Slopehunter.
• Real-world data application to GWAS of subsequent Acute Ischemic Stroke (AIS) or 3pt Major Adverse Cardiovascular Event (MACE) after incident AIS, and Fasting Insulin (FI) adjusted for BMI in the Million Veterans Program (MVP).

Main Results:

• BWMR demonstrated Type I error rates comparable to MR-IVW, MR-WMED, and MR-Egger, but higher than Slopehunter.
• MR-HORSE and Slopehunter showed similar Type I errors for low trait correlations, with MR-HORSE offering better control for large negative correlations and Slopehunter for large positive correlations.
• All evaluated methods exhibited comparable statistical power across different correlation scenarios between incident and subsequent traits.

Conclusions:

• No single index-trait bias correction method is universally optimal across all scenarios.
• Applying multiple bias correction methods and assessing the consistency of results is recommended to identify and correct for index-trait bias effectively.
• The findings suggest a nuanced approach to bias correction in genetic association studies, emphasizing the importance of methodological diversity and cross-validation.