MR-SENSEMAKR

The R package mrsensemakr implements sensitivity analysis tools for Mendelian Randomization, as discussed in Cinelli et al (2020). Robust Mendelian randomization in the presence of residual population stratification, batch effects and horizontal pleiotropy.

Development version

To install the development version on GitHub make sure you have the R package devtools installed. Also make sure to have the latest version of sensemakr (link) installed.

# install.packages("devtools")
devtools::install_github("carloscinelli/sensemakr")
devtools::install_github("carloscinelli/mrsensemakr")

CRAN version coming soon.

Basic usage

## loads package
library(mrsensemakr)

## simulated data example
data("sim_data")

## create vectors indicating variable names in the data
outcome    <- "out.trait" # name of outcome trait
exposure   <- "exp.trait" # name of exposure trait
instrument <- "prs" # genetic instrument (e.g, polygenic risk score)
age.sex    <- c("age", "sex") # age and sex variables (if applicable)
alc.smok   <- c("alcohol", "smoking") # putative pleoitropic vars.
pcs        <- paste0("pc", 1:20) # first 20 principal components pc1 ... pc20

## runs MR sensitivity analysis
mr.sense <- mr_sensemakr(outcome = outcome,
                         exposure = exposure,
                         instrument = instrument,
                         covariates = c(age.sex, alc.smok, pcs), 
                         data = sim_data, 
                         benchmark_covariates = list(alc.smok = alc.smok,
                                                     pcs = pcs))
## print results
mr.sense
#> Sensitivity Analysis for Mendelian Randomization (MR)
#>  Exposure: exp.trait
#>  Outcome: out.trait
#>  Genetic Instrument: prs
#>  Missing Data: No missing data found.
#> 
#> Traditional MR results (2SLS)
#>   MR Estimate (95% CI): 0.807 (0.111 - 1.503)
#>   P-value: 0.02317659
#> 
#> Sensitivity genetic instrument (prs) -> exposure (exp.trait)
#>   Partial R2: 1%
#>   RV (alpha = 0.05): 3.71%
#> 
#> Sensitivity genetic instrument (prs) -> outcome (out.trait)
#>   Partial R2: 0.65%
#>   RV (alpha = 0.05): 1.76%
#> 
#> Bounds on the maximum explanatory power of omitted variables W, if it were as strong as:
#>  bound_label r2zw.x r2dw.zx r2yw.zx adjusted_t_exposure adjusted_t_outcome
#>  1x alc.smok  0.41%   0.05%   0.76%            3.091924            2.34717
#>       1x pcs  2.27%   3.14%   2.51%            2.309587            1.76642


## sensitivity contour plots
plot(mr.sense, 
     benchmark_covariates = list(alc.smok = alc.smok,
                                 pcs = pcs),
     k = list(alc.smok = 1, pcs = 1))

Simulations

Code to reproduce the simulations of the paper can be found in the vignettes folder.