MRPC

MRPC infers directed acyclic graphs (DAGs) representing causal relationships among genotypes and molecular phenotypes by extending the PC algorithm with Mendelian randomization to improve edge orientation in genomic datasets.

Key Features:

• PC algorithm foundation: Extends the PC algorithm for learning directed acyclic graphs (DAGs).
• Improved accuracy: Identifies v-structures (X→Y←Z) and reduces false edges relative to general-purpose graph inference methods.
• Robustness to data arrangement: Produces consistent causal inferences regardless of node arrangement in the input data.
• Mendelian randomization integration: Applies Mendelian randomization using genotype and molecular phenotype data (e.g., gene expression) to constrain and orient edges.
• Versatility for biomedical data: Operates on genomic and other biomedical datasets to provide evidence for causality across multiple data types.

Scientific Applications:

• Gene regulatory network construction: Infers directed gene–gene relationships to construct gene regulatory networks and identify regulatory interactions (X→Y).
• Genotype–phenotype causal inference: Uses genotype and molecular phenotype data to infer causal links relevant to biological processes and disease mechanisms.

Methodology:

Extends the PC algorithm and integrates Mendelian randomization constraints using genotype and molecular phenotype data to identify v-structures and orient edges in causal graphs.