Package 'BWMR'

Bayesian Weighted Mendelian Randomization

Description

Performs variational EM inference for the BWMR model, estimating the causal effect of an exposure on an outcome using summary-level GWAS data.

Usage

BWMR(gammahat, Gammahat, sigmaX, sigmaY, alpha = 100, sqsigma0 = (1e+06)^2)

Arguments

gammahat	Numeric vector of SNP-exposure effect estimates.
Gammahat	Numeric vector of SNP-outcome effect estimates.
sigmaX	Numeric vector of standard errors for the SNP-exposure effects.
sigmaY	Numeric vector of standard errors for the SNP-outcome effects.
alpha	Numeric scalar; concentration parameter for the Beta prior on the proportion of valid instruments. Default is 100.
sqsigma0	Numeric scalar; prior variance for beta. Default is (1e6)^2, representing a diffuse prior.

Value

A list containing:

beta	Causal effect estimate.
se_beta	Standard error of the causal effect estimate.
P_value	P-value from a chi-squared test (1 df).
weights	Posterior mean weights for each SNP.
tau	Estimate of the pleiotropy noise parameter.
sigma	Estimate of the instrument strength parameter.
mu_pi	Posterior mean of the proportion of valid instruments.
plot1	Scatter plot of SNP effects with error bars (ggplot2).
plot2	ELBO convergence plot (ggplot2).
plot3	Bar chart of posterior mean weights per SNP (ggplot2).
plot4	Weighted scatter plot with regression line (ggplot2).

Examples

data(ExampleData)
result <- BWMR(
  gammahat = ExampleData$beta.exposure,
  Gammahat = ExampleData$beta.outcome,
  sigmaX   = ExampleData$se.exposure,
  sigmaY   = ExampleData$se.outcome
)
result$beta
result$P_value

---

Bayesian Weighted Mendelian Randomization (no plots)

Description

A faster version of BWMR() that omits diagnostic plots. Suitable for simulation studies or large-scale analyses where plotting is unnecessary.

Usage

BWMR_quick(
  gammahat,
  Gammahat,
  sigmaX,
  sigmaY,
  alpha = 100,
  sqsigma0 = (1e+06)^2
)

Arguments

gammahat	Numeric vector of SNP-exposure effect estimates.
Gammahat	Numeric vector of SNP-outcome effect estimates.
sigmaX	Numeric vector of standard errors for the SNP-exposure effects.
sigmaY	Numeric vector of standard errors for the SNP-outcome effects.
alpha	Numeric scalar; concentration parameter for the Beta prior on the proportion of valid instruments. Default is 100.
sqsigma0	Numeric scalar; prior variance for beta. Default is (1e6)^2, representing a diffuse prior.

Value

A list containing:

beta	Causal effect estimate.
se_beta	Standard error of the causal effect estimate.
P_value	P-value from a chi-squared test (1 df).
tau	Estimate of the pleiotropy noise parameter.
sigma	Estimate of the instrument strength parameter.
mu_pi	Posterior mean of the proportion of valid instruments.

Examples

data(ExampleData)
result <- BWMR_quick(
  gammahat = ExampleData$beta.exposure,
  Gammahat = ExampleData$beta.outcome,
  sigmaX   = ExampleData$se.exposure,
  sigmaY   = ExampleData$se.outcome
)
result$beta
result$P_value

---

Example Mendelian randomization genotype summary level dataset

Description

An example Mendelian randomization dataset.

Usage

ExampleData

Format

ExampleData

A data frame with 76 rows and 5 columns:

SNP

SNP genotype

beta.exposure

SNP-exposure estimates

beta.outcome

SNP-outcome estimates

se.exposure

Standard errors of the SNP-exposure estimates

se.outcome

Standard errors of the SNP-outcome estimates

---