A scalable simulation tool for epidemiological studies
EpiGraph is a scalable, fully distributed simulator that is able to perform large scale and realistic stochastic simulations of the propagation of the flu virus. EpiGraph source code is available. Now you can download, modify and execute EpiGraph using our data (from 1 up to 92 urban areas, from 1 up to 130 processes). In addition, we provide complementary material for creating new urban areas. Go to the download section to access to the source code and user manual.
- Why use EpiGraph?. The use of EpiGraph in interesting from two different perspectives. First, it can be used by epidemiologist to perform large-scale simulations of influenza propagation. Second, it can be used as a parallel benchmark to evaluate new architectures or optimization techniques. From the computational perspective, EpiGraph is a network I/O bound application written in C language and implemented on MPI. Internally, EpiGraph performs irregular memory access patterns related to sparse matrices used to model the individual interactions.
- How to cite EpiGraph?. If you are using EpiGraph and you want to cite it, please, cite the following reference: Gonzalo Martin, David E. Singh, Maria-Cristina Marinescu and Jesus Carretero. Towards efficient large scale epidemiological simulations in EpiGraph. Parallel Computing. Vol. 42, No. 0, Pages: 88-102. 2015.
The current implementation of EpiGraph includes functionalities for modeling via a realistic interconnection network based on actual individual interactions extracted from social networks and demographical data. This network includes the geographic location of each individual and a transportation model which allows the study of the spatial dynamics of the virus propagation over-large scale areas. EpiGraph also includes a model of the interaction between influenza spreading and climatic and meteorological factors, such as temperature, atmospheric pressure and humidity levels.
In the following figure we can observe different stages of the simulation of an outbreak of influenza for the 92 most populated cities in Spain assuming that it is originated in A Coruña. By means of EpiGraph it is possible to assess the impact of the infection on the population and the temporal evolution of the propagation at individual level including local data such as specific environment conditions and social data. In addition, EpiGraph is currently used to analyse the dissemination of the infection and the effect of different policies (like vaccination or quarantine) on the progress of the spread.
- Epidemic model design. This model captures the mechanism by which susceptible individuals get infected and go through the different stages of the infection. This model is specific to the infectious agent under study, in our case, to the influenza virus. At the level of the individual, we allow for the modeling of characteristics such as age, gender, and race. We have implemented a detailed modelling of the influenza virus consisting of 17 different states of the infection including latent, presymptomatic, asymptomatic, hospitalized and under vaccination.
- Social model design. We developed a social model for the population under study, including the patterns of contact between individuals within this population. The social model is represented via an undirected connection graph and can capture heterogeneity features at the level of both the individual and each of his/her interactions. We use contact information from social networks to realistically approximate these connections. We allow for customizing individual interaction behavior based on the day of the week and the time of day. In this way, we model social structures such as companies, schools, or groups of stay-at-home parents and retired people that are interacting in education programs, hobby classes, kids’ schools or any other kind of activities that make them come in contact with each other. In the following figure we can see an example of a reduced contact network used in EpiGraph.
- Environment design. An important project research line is to enhance the social model with a detailed environment design. With the implementation of the Internet of Things, more and more elements of our common life can be tracked and processed. This data provides very valuable information that can be used for increasing the accuracy of EpiGraph. For instance, in urban environments (Smart Cities) traffic, public transport and pedestrian conditions can be used to model social interactions. Data collected from hospitals or pharmacies will contribute to model the spread of the epidemic. A more detailed analysis of social networks allows for the consideration of geolocalization data as well as providing additional information about the number of infected individuals. Finally, the use of meteorological forecasts and air pollution levels (Smart Environments) allows for improved simulation given that these values affect the ability of the virus to be propagated.
- Distributed and scalable simulator. We implement a scalable, fully distributed simulator in MPI. Currently EpiGraph can be executed using tens compute nodes and can perform simulations of a complete season for environments with hundreds of millions inhabitants.
- Realistic simulations. We have validated the results of the simulations with other simulators as well as real data obtained from NYSDOH. In all the cases EpiGraph is able to provide an accurate prediction of the epidemic propagation.
- Study of different scenarios. We investigate the virus dissemination process and compare it with dissemination in networks which have exponential and normal contact distributions, as well as in a social model without time-dependent interactions. We also study how infecting different types of individuals may affect the epidemic. In addition, we analyse and compare the impact of different vaccination policies on managing the virus dissemination process.
- At Carlos III University of Madrid: David E. Singh, Jesús Carretero and Florin Isaila.
- At National Centre for Epidemiology, Carlos III Institute of Health and CIBER en Epidemiología y Salud Pública (CIBERESP): Amparo Larrauri, Diana Gomez-Barroso and Concepción Delgado-Sanz
- At Barcelona Supercomputing Center: María-Cristina Marinescu
- At industry: Gonzalo Martín
This work has been partially funded by the Spanish Ministry of Science and Education under the MEC 2011/00003/001 and TIN2010-16497 contracts. We also would like to thank Agencia Estatal de Meteorología (AEMET) for providing meteorological data of Spain.