CausalDAG

CausalDAG infers differences between causal directed acyclic graphs (DAGs) representing gene regulatory networks to identify edges that appear, disappear, or change in weight across biological conditions.

Key Features:

• Difference Causal Inference (DCI) algorithm: Identifies changes in gene regulatory mechanisms between two conditions by inferring differences between causal DAGs.
• Efficiency and scalability: Performs sample-efficient inference by directly estimating differences between two causal graphs instead of separately estimating each large causal graph.
• Robustness via stability selection: Supports stability selection across tuning parameters to obtain consistent difference causal graphs.
• Versatility in data application: Accepts bulk and single-cell RNA-seq expression data from diverse conditions and cell states.
• Intervention prediction: Predicts effects of interventions within inferred gene regulatory networks.

Scientific Applications:

• Genomics and systems biology: Enables modeling and comparison of regulatory network structure across biological states.
• Disease mechanism discovery: Facilitates identification of regulatory changes associated with disease states.
• Developmental process analysis: Supports detection of regulatory differences during developmental transitions.
• Design of experiments and therapies: Informs targeted experiment design and therapeutic strategy development via predicted intervention effects.

Methodology:

Apply the DCI algorithm to large-scale expression datasets (bulk or single-cell RNA-seq) from two conditions to infer difference DAGs directly rather than estimating each DAG separately, optionally using stability selection across tuning parameters; the methods are implemented in Python.