Novel methods for epidemic prediction

Title of the research project

SIBYL - Statistical Inference via Belief Propagation for Dynamical Models of Epidemics

Scientific area

Statistical inference, Complex networks, Statistical methods for epidemiology

Project coordinator

Alfredo Braunstein

Abstract  

Epidemic outbreaks are generally discovered when a large portion of the population is already infected. Promptly identifying the origin of the infection can provide invaluable information to fight the disease. SIBYL project will develop practical mathematical and computational solutions for this and many related problems.

Description of the research project 

Networks have been used to model infectious disease spreading in different forms. In general, the nodes of the network represent individuals and the edges describe the social contacts through which the infection can spread. The detailed structure and temporal evolution of contact networks were inaccessible until not long ago, but this is rapidly changing due to the diffusion of online social networks on mobile phones and the miniaturization of radio-frequency identification devices. Access to such information, together with an observation of the state of infection at a given time (the present), allows to mathematically pose and solve a variety of Bayesian inference problems, including an estimation of past history of an outbreak (e.g. the “patient-zero” problem, i.e. the identification of the first infected individual) and a much more detailed prediction of its future evolution (e.g. the forecast of the total number of affected individuals and the time duration of the outbreak). The problem is mathematically particularly difficult because both input ingredients of the inference problem, i.e. the propagation network and the infection observation, are typically noisy or inaccurate. The project will address moreover the problem of completing and correcting our knowledge of the underlying network and the identification of external contagion channels by integrating information from several infection outbreaks.

Impact on society 

Efficient forecast of epidemics, the identification of the key spreaders and undetected contact channels, are fundamental issues in the design of effective containing strategies, which can have a huge economical and social impact. Moreover, it is expected that SIBYL will raise further interest on collecting data on contacts through which infection can propagate, especially in controlled environments such as critical-care hospital wards and livestock.

Project results

Studying the random structure which determinates epidemic history with a mathematical model has allowed the deduction of its future development. Model parameters have been determined through statistical inference method, including the ones related to infectiousness. The study has been applied to bovine epidemics analysis and has allowed the development of algorhitms and computational methods for non specialists.

In particular:

  • 1 publication on PNAS (Proceedings of the National Academy of Sciences)
  • New research collaborations among them Ecole Normale Superieure
  • Participation to the Horizon 2020 MSCA - RISE  INFERNET project with the l’Italian Institute for Genomic Medicine (Torino, Italy)
  • Collaboration with  Istituto Zooprofilattico Sperimentale della Lombardia e dell'Emilia Romagna
  • 1 research position activated

Working group 

At Politecnico di Torino:

Luca dall’Asta, Andrea Pagnani

  • Budget: 110.935
  • Start date: 15/12/2015
  • End date: 14/12/2017