MRBEE

The MRBEE package is designed for conducting multivariable Mendelian Randomization (MVMR) analyses. MRBEE.IMRP removes weak instrument bias, which is caused by the estimation error of exposures and outcome GWAS, by using an unbiased estimatin function. On the other hand, MRBEE iteratively detected and remove the horiztonal pleiotropy using a pleiotropy test, making it robust to horizontal pleiotropy.

Installation

You can install the MRBEE.IMRP package directly from GitHub using the following command:

# Install the devtools package if you haven't already
if (!requireNamespace("devtools", quietly = TRUE))
  install.packages("devtools")

# Install MRBEE.IMRP from GitHub
devtools::install_github("noahlorinczcomi/MRBEE")

Additionally, one could utilize the ldscR package for estimation of the covariance matrix of estimation errors. You can install ldscR from GitHub as well:

# Install ldscR from GitHub
devtools::install_github("harryyiheyang/ldscR")

Usage

Here's an example workflow using MRBEE.IMRP and ldscR:

Data Preparation and Harmonization

First, download and prepare your GWAS summary data. Then, harmonize alleles using the filter_align() function:

data_url <- "http://tinyurl.com/nhdfwd8v"
temp_file <- tempfile()
download.file(data_url, temp_file, mode="wb")
gwaslist=readRDS(temp_file)
unlink(temp_file)
library(MRBEE)
data("hapmap3")
gwaslist=filter_align(gwas_data_list=gwaslist,ref_panel=hapmap3[,c("SNP","A1","A2")])

Estimation Error Covariance Matrix

ZMatrix=cbind(gwaslist$driving$Zscore,gwaslist$computer$Zscore,gwaslist$TV$Zscore,gwaslist$schooling$Zscore,gwaslist$myopia$Zscore)
Rxy=errorCov(ZMatrix=ZMatrix)

While both ldscR and the insignificant GWAS effect estimation method can be used for estimating the covariance matrix, slight differences exist between the two approaches.

#library(ldscR)
#data("EURLDSC")
#fitldsc=ldscR(GWAS_List=gwaslist,LDSC=EURLDSC)
#Rxy=fitldsc$ECovEst

However, these differences do not significantly impact the final results.

Perform Joint test to detect significant IVs

jointtest=Joint.test(bZ=ZMatrix[,-5],RZ=Rxy[-5,-5])
jointtest$SNP=gwaslist$driving$SNP
########################### Perform C+T using PLINK ########################################
#write.table(jointtest,"myopia/plinkfile/joint.txt",row.names=F,quote=F,sep="\t")
#setwd("~/Plink")
#system("./plink --bfile data/1000G/1kg_phase3_EUR_only --clump myopia/plinkfile/joint.txt --clump-field P  --clump-kb 500 --clump-p1 5e-8 --clump-p2 5e-8 --clump-r2 0.01 --out myopia/plinkfile/joint")
#plink=fread("~/myopia/plinkfile/joint.clumped")[,1]
plink=data.table::fread("http://tinyurl.com/8yt7bhvp", header = F)

Perform MVMR using Zscore

We recommend using Z-scores for MR analysis. If effect sizes are estimated with the same sample size, Z-scores and effect sizes are equivalent. However, for effect sizes estimated from smaller sample sizes, Z-scores naturally assign them less weight in estimating causal effects. This resembles a second-stage reweighting, leading to more stable causal estimates.

ZMatrix1=ZMatrix[which(jointtest$SNP%in%plink$V1),]
fit=MRBEE.IMRP(by=ZMatrix1[,5],bX=ZMatrix1[,-5],byse=rep(1,nrow(ZMatrix1)),bXse=matrix(1,nrow(ZMatrix1),4),Rxy=Rxy,var.est="ordinal")
print(fit$theta/sqrt(diag(fit$covtheta)))

Perform MVMR using Effect Size and Standard Error

BETA=cbind(gwaslist$driving$BETA,gwaslist$computer$BETA,gwaslist$TV$BETA,gwaslist$schooling$BETA,gwaslist$myopia$BETA)
BETA=abs(BETA)*sign(ZMatrix) # The filter_align function only adjust the signs of Zscore
SE=cbind(gwaslist$driving$SE,gwaslist$computer$SE,gwaslist$TV$SE,gwaslist$schooling$SE,gwaslist$myopia$SE)
BETA1=BETA[which(jointtest$SNP%in%plink$V1),]
SE1=SE[which(jointtest$SNP%in%plink$V1),]
fit1=MRBEE.IMRP(by=BETA1[,5],bX=BETA1[,-5],byse=SE1[,5],bXse=SE1[,-5],Rxy=Rxy,var.est="ordinal")
print(fit1$theta/sqrt(diag(fit1$covtheta)))

Maintainers

This package is maintained by:

Yihe Yang Email: yxy1234@case.edu ORCID: 0000-0001-6563-3579

Noah Lorincz-Comi Email: njl96@case.edu ORCID: 0000-0002-0517-2499

remlapmot/MRBEE