MR-BMA
This function allows the user to pass in a set of harmonized genetic
variants generated using the
TwoSampleMR::mv_harmonise_data() function, and outputs the
results of a Mendelian Randomization Bayesian Model Averaging
experiment.
# Extract genetic instruments
lipid_exposures <- TwoSampleMR::mv_extract_exposures(id_exposure = c("ieu-a-299", "ieu-a-300", "ieu-a-302"))
#> API: public: http://gwas-api.mrcieu.ac.uk/
#> Please look at vignettes for options on running this locally if you need to run many instances of this command.
#> Clumping 1, 214 variants, using EUR population reference
#> Removing 70 of 214 variants due to LD with other variants or absence from LD reference panel
#> Extracting data for 144 SNP(s) from 3 GWAS(s)
#> Harmonising HDL cholesterol || id:ieu-a-299 (ieu-a-299) and LDL cholesterol || id:ieu-a-300 (ieu-a-300)
#> Harmonising HDL cholesterol || id:ieu-a-299 (ieu-a-299) and Triglycerides || id:ieu-a-302 (ieu-a-302)
# Extract corresponding outcome data
cad_outcome <- TwoSampleMR::extract_outcome_data(snps = lipid_exposures$SNP, outcomes = "ebi-a-GCST005195")
#> Extracting data for 144 SNP(s) from 1 GWAS(s)
#> Finding proxies for 2 SNPs in outcome ebi-a-GCST005195
#> Extracting data for 2 SNP(s) from 1 GWAS(s)
# Generate harmonized dataset
lipids_cad_harmonized <- TwoSampleMR::mv_harmonise_data(lipid_exposures, cad_outcome)
#> Harmonising HDL cholesterol || id:ieu-a-299 (ieu-a-299) and Coronary artery disease || id:ebi-a-GCST005195 (ebi-a-GCST005195)
#> Removing the following SNPs for being palindromic with intermediate allele frequencies:
#> rs7534572
# Run MR-BMA
mr_bma_res <- mr_bma(lipids_cad_harmonized, calculate_p = TRUE, nrepeat = 1000)
#> ℹ Preparing input
#> ✔ Preparing input [64ms]
#>
#> ℹ Running MR-BMA
#> ✔ Running MR-BMA [49ms]
#>
#> ℹ Removing influential variants
#> ✔ Removing influential variants [106ms]
#>
#> ℹ Removing outliers
#> ✔ Removing outliers [115ms]
#>
#> ℹ Estimating empirical p-value using 1000 permutations
#> ✔ Estimating empirical p-value using 1000 permutations [3.2s]
#>
#> ℹ Performing Nyholt correction for effective number of tests
#> ✔ Performing Nyholt correction for effective number of tests [31ms]
#>
# Output best models
mr_bma_res$model_best
#> # A tibble: 7 × 3
#> rf_combination posterior_prob causal_estimate
#> <chr> <dbl> <chr>
#> 1 ieu-a-300,ieu-a-302 0.843 0.365,0.219
#> 2 ieu-a-299,ieu-a-300,ieu-a-302 0.156 -0.063,0.361,0.19
#> 3 ieu-a-299,ieu-a-300 0 -0.131,0.376
#> 4 ieu-a-300 0 0.393
#> 5 ieu-a-302 0 0.311
#> 6 ieu-a-299,ieu-a-302 0 -0.105,0.26
#> 7 ieu-a-299 0 -0.202
# Output marginal inclusion probabilities for each risk factor
mr_bma_res$mip_table
#> # A tibble: 3 × 6
#> exposure mace mip p_val p_fdr p_nyh…¹
#> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 ieu-a-299 -0.00987190503160796 0.15651065273458 0.924075924075… 0.92… 1
#> 2 ieu-a-300 0.364151410155237 1 0.000999000999… 0.00… 0.0019…
#> 3 ieu-a-302 0.214705609094214 0.999726011149142 0.000999000999… 0.00… 0.0019…
#> # … with abbreviated variable name ¹p_nyholt