PyMR: Mendelian Randomization in Python

A modern, test-driven Python package for Mendelian Randomization (MR) analysis.

Features

• Multiple MR methods: IVW, weighted median, MR-Egger, mode-based
• Bayesian MR: Full posterior inference, Bayes factors, model comparison
• Sensitivity analyses: Heterogeneity tests, MR-PRESSO, leave-one-out
• Data harmonization: Automatic allele alignment and strand flipping
• GWAS integration: Load from Pan-UKB, IEU OpenGWAS, or custom files
• Visualization: Forest plots, scatter plots, funnel plots, posterior distributions
• No external dependencies: Pure Python (no PLINK, PyMC, or Stan required)

Installation

pip install pymr

Quick Start

Frequentist MR

from pymr import MR, load_gwas

# Load GWAS summary statistics
exposure = load_gwas("bmi_gwas.tsv.gz")
outcome = load_gwas("diabetes_gwas.tsv.gz")

# Run MR analysis
mr = MR(exposure, outcome)
results = mr.run()

print(results)
#              method      beta        se        OR      pval  nsnp
# 0               IVW  0.924740  0.030497  2.521214  5.83e-202   192
# 1   Weighted Median  1.039266  0.021565  2.827140  0.00e+00    192
# 2          MR-Egger  0.911587  0.075401  2.488269  1.19e-33    192

Bayesian MR

from pymr import BayesianMR

# Run Bayesian MR with full posterior inference
bmr = BayesianMR(harmonized_data, prior_mean=0, prior_sd=1)
bmr.sample(n_samples=10000, n_chains=4, warmup=1000)

# Get posterior summary
summary = bmr.summary()
print(f"Effect: {summary['mean']:.3f} [{summary['ci_lower']:.3f}, {summary['ci_upper']:.3f}]")
print(f"Bayes Factor: {bmr.bayes_factor(null_value=0):.2f}")

# Visualize posterior
bmr.plot_posterior()

See docs/bayesian_mr.md for comprehensive Bayesian MR documentation.

Development

# Clone repository
git clone https://github.com/maxghenis/pymr.git
cd pymr

# Install in development mode
pip install -e ".[dev,docs]"

# Run tests
pytest

# Build documentation
jupyter-book build docs

Citation

If you use PyMR in your research, please cite:

@software{pymr2025,
  author = {Ghenis, Max},
  title = {PyMR: Mendelian Randomization in Python},
  year = {2025},
  url = {https://github.com/maxghenis/pymr}
}

License

MIT License - see LICENSE for details.