Static graph approximations of dynamic contact networks for epidemic forecasting

Area of Science:

• Epidemiology
• Network Science
• Computational Biology

Background:

• Epidemic modeling is crucial for understanding infectious disease spread and informing control strategies.
• Network-based models integrating contact graphs show promise but face challenges with unobserved future network patterns.
• Current methods using full static graphs can overestimate epidemic characteristics.

Purpose of the Study:

• To develop novel static network approximation methods for epidemic modeling.
• To address the overestimation issues associated with the standard full static graph method.
• To improve the accuracy of epidemic forecasting by preserving network sparsity.

Main Methods:

• Proposed two novel static network approximation methods: DegMST and EdgeMST.
• DegMST utilizes the frequency of temporal edges, while EdgeMST uses node degrees to preserve network sparsity.
• Evaluated methods on seven real-world contact networks.

Main Results:

• DegMST and EdgeMST more accurately represent dynamic graph characteristics than the full static graph.
• EdgeMST demonstrated superior accuracy in estimating disease dynamics for epidemic forecasting.
• The proposed methods better preserve the sparsity of real-world contact networks.

Conclusions:

• DegMST and EdgeMST offer improved static network approximations for epidemic modeling.
• EdgeMST provides more accurate epidemic forecasting compared to the standard full static method.
• Preserving network sparsity is key to enhancing the reliability of epidemic models.