FIDDLE

FIDDLE extracts knowledge from published scientific literature to assemble or extend computational network models for systems and computational biology.

Key Features:

• Automated Knowledge Extraction: Uses machine reading techniques to extract relevant interactions and statements from published literature for model integration.
• Diagram-Driven Approach: Employs network diagrams as foundational structures for model assembly and extension.
• Algorithmic Innovations: Implements Breadth First Addition (BFA), which explores candidate edges in a breadth-first manner prioritizing specificity by focusing on abundant interactions (even if mixed with false information), and Depth First Addition (DFA), which uses a depth-first strategy emphasizing sensitivity by integrating fewer but higher-confidence interactions.
• Comprehensive Evaluation: Algorithms were evaluated on Erdös-Rényi random networks (ER) and Barabási-Albert scale-free networks (BA) as proxies for biological networks to assess reconstruction of specified network structures and replication of behaviors of predefined "golden models."

Scientific Applications:

• Systems Biology Model Updating: Automates integration of literature-derived interactions into existing models to support model updating and extension in systems biology.
• Network Reconstruction and Validation: Supports reconstruction of network structures and replication of predefined "golden model" behaviors for model validation studies.

Methodology:

FIDDLE applies BFA and DFA algorithms to baseline models with varying candidate edges and evaluates performance on ER and BA networks by assessing network reconstruction and replication of "golden model" behaviors, noting limited success on scale-free (BA) networks due to redundancy and complexity.