MRCIP

MRCIP models correlated and idiosyncratic pleiotropy to enable causal effect estimation in Mendelian randomization using genetic variants as instrumental variables.

Key Features:

• Instrumental variable-based causal estimation: Leverages genetic variants as instrumental variables to estimate the causal effect of an exposure on an outcome while accounting for unmeasured confounding.
• Random-effect model for correlated pleiotropy: Explicitly models correlated pleiotropy via a random-effect framework.
• Weighting scheme for idiosyncratic pleiotropy: Applies a novel weighting scheme to manage variant-specific (idiosyncratic) pleiotropic effects.
• PRW-EM algorithm: Estimates model parameters using a Penalized Reweighted Expectation-Maximization (PRW-EM) algorithm.
• Weighted likelihood maximization: PRW-EM maximizes a weighted likelihood function tailored to pleiotropic complexities.
• Quantification and testing: Estimates the degree of correlated pleiotropy and performs a likelihood ratio test to assess its presence.

Scientific Applications:

• Causal inference in genetic epidemiology: Estimate causal effects of exposures on outcomes using Mendelian randomization with genetic instrumental variables.
• Robustness under pleiotropy: Improve robustness of MR causal estimates when correlated and idiosyncratic pleiotropy are present.
• Pleiotropy assessment: Quantify and test for the presence and degree of correlated pleiotropy using likelihood ratio testing.

Methodology:

MRCIP implements a random-effect model for correlated pleiotropy, a weighting scheme for idiosyncratic pleiotropy, parameter estimation via a Penalized Reweighted Expectation-Maximization (PRW-EM) algorithm that maximizes a weighted likelihood, and a likelihood ratio test to assess correlated pleiotropy.