CARNIVAL

CARNIVAL infers context-specific signaling networks from gene expression data using causal reasoning and integer linear programming to identify upstream regulators and network topology.

Key Features:

• Integration of Prior Knowledge: Incorporates signed and directed protein-protein interactions, transcription factor (TF) targets, and pathway signatures as prior biological knowledge.
• Inference of Transcription Factor Activities and Pathway Scores: Uses DoRothEA to infer TF activities and PROGENy to derive pathway scores from gene expression data.
• Integer Linear Programming (ILP) Framework: Formulates network identification as an ILP solved with solvers such as CPLEX to balance fitting error and model size.
• Efficient Computation: The ILP formulation enables efficient computation suitable for large-scale network inference.
• Incorporation of Drug Targets: Integrates drug-target information alongside protein-protein interaction networks and pathway scores when constructing inference constraints.

Scientific Applications:

• Signaling network contextualization from transcriptomics: Contextualizes signaling networks from gene expression data to reveal upstream processes driving transcriptional changes.
• Disease mechanism discovery (IgA nephropathy): Applied to IgA nephropathy (IgAN) to identify dysregulated pathways such as Wnt and TGF-β and generate experimentally validated hypotheses.

Methodology:

Infers TF activities and pathway scores from gene expression using DoRothEA and PROGENy, constructs linear constraints from inferred values together with protein-protein interaction networks, drug targets, and pathway scores, and formulates an ILP solved with a solver like CPLEX to identify sub-network topologies that fit the data while minimizing model complexity.