Active querying approach to epidemic source detection on contact networks

Area of Science:

• Epidemiology and network science.
• Computational epidemiology and infectious disease modeling.

Background:

• Identifying the source of an epidemic (patient zero) is crucial for containment.
• Traditional methods often require observing many individuals, which is not always feasible.
• Existing approaches may not account for unknown epidemic duration.

Purpose of the Study:

• To develop an active querying strategy for identifying epidemic sources when observations are limited.
• To address the challenge of unknown epidemic duration by incorporating a Bayesian prior.
• To propose methods for selecting informative individuals to query for their infection status.

Main Methods:

• Formulated an active querying problem alternating between source inference and information-gathering steps.
• Employed a Bayesian approach for source inference, including a prior on epidemic duration.
• Developed querying strategies inspired by active learning, focusing on individuals maximizing prediction disagreement.

Main Results:

• Active querying significantly improves epidemic source inference compared to baseline heuristics.
• A strategy maximizing disagreement between individual and consensus predictions proved most effective.
• Effective source inference was achieved by querying only a small fraction of individuals.

Conclusions:

• Active querying is a powerful approach for identifying epidemic sources with limited observations.
• The proposed method is applicable to both static and temporal contact networks.
• This strategy offers a practical and efficient way to enhance epidemic source detection in real-world scenarios.