MarginalCausality

MarginalCausality infers marginal causal effects from observational and interventional transcriptomic data to identify downstream causal relationships of a single knocked-out gene using Gaussian directed acyclic graphs (DAGs).

Key Features:

• Marginal Causal Estimation: Employs a marginal causal estimation method based on Gaussian directed acyclic graphs (DAGs) to infer causal relationships between a knocked-out gene and many other genes.
• Observational and Interventional Integration: Integrates observational and interventional (single-gene knockout) data to distinguish downstream causal effects from upstream or associative interactions.
• Simulation Study Validation: Validated through simulation studies that demonstrate the ability to differentiate types of genetic interactions and reliably estimate total causal effects.
• Comparison with Differential Analysis: Performs comparably to classical differential analysis methods when biological replicates are abundant while providing formal causal interpretations.
• Scalability and Computational Efficiency: Enables simultaneous analysis of several thousands of genes to mitigate parameter-estimation challenges in large transcriptomic datasets.
• Gene Prioritization: Identifies subsets of genes for focused causal network analysis and further investigation.

Scientific Applications:

• Transcriptomic Knockout Studies: Inferring downstream causal effects in transcriptomic experiments that include single-gene knockouts using combined observational and interventional data.
• Gene Interaction Characterization: Prioritizing genes and characterizing causal interactions underlying biological processes and diseases for follow-up network analyses.

Methodology:

Marginal causal estimation within Gaussian directed acyclic graphs (DAGs) using observational and interventional (single-gene knockout) data, validated via simulation studies.