Provenance Tracing in Network Diffusion Algorithms.

Area of Science:

• Systems Biology
• Network Biology
• Computational Biology

Background:

• Diffusion algorithms are widely used in biological applications.
• Provenance tracing for these algorithms has been overlooked.
• Understanding information propagation in biological networks is crucial.

Purpose of the Study:

• To propose a novel path-based strategy for provenance tracing in random walk-based network diffusion algorithms.
• To introduce quantitative measures for evaluating diffusion effectiveness and node importance.
• To apply and validate the framework on biological networks, including SARS-CoV-2 interactors.

Main Methods:

• Developed a path-based approach for ranking path contributions to node scores.
• Introduced 'path-based effective diffusion' and 'diffusion betweenness' metrics.
• Applied the framework to Regularized Laplacian and Random Walk with Restart algorithms on human PPI and signaling networks.

Main Results:

• Demonstrated that diffusion algorithms utilize multi-edge paths and network topology.
• Identified key proteins via diffusion betweenness, enriched with essential human genes and viral interactors.
• Showed overlap between top contributing paths and COVID-19 pathways in a human signaling network.

Conclusions:

• The path-based framework provides valuable insights into diffusion algorithms.
• Offers a powerful tool for biologically meaningful interpretation of diffusion scores.
• Complements existing systems biology approaches by providing a path-centric perspective.