Simulation studies for comparing methods for Mendelian randomization

Individual-level simulations

Summary statistics are generated by simulating individual-level genotype and phenotype data followed by association analysis by linear regression.

Located in folder individual_level_simulations:

• bal_InSIDE_simulate_beta.R: Individual-level simulations under balanced pleiotropy and InSIDE assumption satisfied.
• bal_noInSIDE_simulate_beta.R: Individual-level simulations under balanced pleiotropy and InSIDE assumption violated.
• merge_individual_betahat.R: Merge data from all simulation groups.
• MR_analysis.R: Conduct MR analysis on simulated summary statistics.

Summary-level simulations

Summary statistics are simulated directly from the summary-level model implied by the individual-level model without generating individual genotype and phenotype data. Summary level simulations are substantially faster than individual-level simulations.

Located in folder summary_level_simualtions:

• bal_InSIDE_smry.R: Summary-level simulations under balanced pleiotropy and InSIDE assumption satisfied.
• bal_noInSIDE_smry.R: Summary-level simulations under balanced pleiotropy and InSIDE assumption violated.
• dir_InSIDE_smry.R: Summary-level simulations under directional pleiotropy and InSIDE assumption satisfied.
• dir_noInSIDE_smry.R: Summary-level simulations under directional pleiotropy and InSIDE assumption violated.

Function for MR-Lasso implementation

MR_lasso.R: function to implement MR-Lasso, obtained from the following paper: Slob, Eric AW, and Stephen Burgess. "A Comparison Of Robust Mendelian Randomization Methods Using Summary Data." BioRxiv (2019): 577940.