Data Driven Network Modelling for Epidemiology in Dynamic Human Networks

Lead Research Organisation: University of Cambridge
Department Name: Computer Laboratory

Abstract

My research will develop data-driven modelling of human interaction dynamics, where experimental measurements are followed by mathematical modelling. I emphasise that real-world data needs to drive modelling. Such refined modelling will predict potential disease outbreaks and enables building synthetic networks, which will provide opportunities to scale up the network environment and experimentally control epidemics. I aim to build such a prediction system. Realising this vision will involve both sophisticated data collection and model construction. Especially data collection takes an important role. Current popular detection mechanisms using WiFi access points or short range radio involve high failures, communication protocol limitation and complex statistics. Without in-depth understanding of data collection mechanisms, modelling such networks will not be reliable. The derived epidemic models will need to be accurate and parameterised with data on human interaction patterns, modularity, and details of time dependent activity. Thus, a model can determine epidemic spread accordingly, and synthetic networks can be constructed. Data collection will also require careful attention on ethics and privacy issues. The epidemic spread of infectious disease remains a serious threat, both in the UK and around the world. The proposed research aims at understanding human interactions in the real world using wireless technology to develop advanced modelling of epidemic spread. This research will provide unique insight into medical and statistical problems, ultimately contributing to diminish control epidemic spread at the public health level. Infectious disease epidemics are analogous to wireless computer epidemics, especially when computer devices (e.g. mobile phones) are carried by people, as both types of epidemics rely on human interactions. I propose to advance research in both infectious disease epidemiology and in wireless computer epidemiology in two ways. First, I will aid our understanding of social networks by extending and developing analysis and modelling approaches with empirical real-world human connectivity data. Second, I will establish high quality data collection by investigating effective approaches using multiple hardware and communication mechanisms. The proposed research will provide an advanced model of epidemic spread of infectious disease, and the results will highlight solutions to medical and statistical issues which could not be addressed before. The outcome of this effort will provide more accurate and improved predictions of infectious disease epidemics.The quantitative understanding of human interactions is complex and has not been explored at any depth. Theoretical modelling and simulation based approaches are limited, and rich real-world data will be key to refine the modelling. Current models in network theory are too simplified, and multiple large-scale experimental data are needed both for modelling and building systems. The recent emergence of wireless technology provides a unique opportunity to collect precise human connectivity data. For example, people can carry tiny wireless sensors (<1 cm^2) that record dynamic information about other devices nearby. A post-facto analysis of this data will yield valuable insight into complex human interactions, which in turn will support meaningful modelling of understanding networks. Specific individuals can be identified who act as coalescing hubs at different points in space and time and who influence data flow. By neutralising such hubs, we can prevent the spread of viruses. The developed prediction system in the proposed research can be used in various ways. Sexually transmitted diseases are on the increase: the AIDS epidemic of the past two decades is a prime example of a situation that could be stopped by the prediction system before reaching the fatal stage.

Publications

10 25 50
 
Description The project 'Data Driven Network Modelling for Epidemiology in Dynamic Human Networks (DDEPI)' (2010-2015) has developed the FluPhone project in 2010 (http://www.cl.cam.ac.uk/research/srg/netos/fluphone2/), which aims at collecting human proximity information from the general population for building time dependent contact networks. Data collection for symptoms from infectious disease is built over the framework of the Haggle project (http://www.haggleproject.org/ 2006-2010). The FluPhone project has shown another dimension of human contact networks study. The project has been reported in many news media, including University of Cambridge Press (http://www.cam.ac.uk/research/features/fluphone-disease-tracking-by-app), BBC (http://www.bbc.co.uk/news/uk-england-cambridgeshire-13281131), Times newspaper, and others.

In 2013-2014, the project evolved to understand infectious diseases in Africa. To build decentralised Raspberry Pi networks together with RFID tags, we have designed the Raspberry Pi OpenBeacon Reader networks. This network makes it possible to collect human mobility data, where the Internet access is not deployed, such as in rural areas in Africa. Obtaining mobility data at schools is valuable, as these ore usually mixing places of different families and interaction data can be used for understanding the spread of infectious diseases.

In 2014-15, the EpiMap project has been explored for a system of opportunistic networks combined with satellite communication, designed to face the challenges posed by weak power and communications infrastructure in rural regions of developing countries in Asia, Africa and South America. I will use a delay-tolerant small satellite for data transfer between developing countries and Europe and North America. Data collected through EpiMap can be used to help design more efficient vaccination strategies and equitable control programmes.

In 2015, unique experiments of human connectivity measurement and communication networking has been deployed. RFID tags and GPS are used to measure human proximity in the Brazilian Amazon rural region, without constant power supply or Internet. Small river communities are fragmented and the only transportation is by boat. This information will be used to develop improved mathematical models of the spread of infectious diseases. The modelling is complemented by a survey to understand the living conditions in the relevant rural communities. The outcome of the project will be potentially used to help design more efficient vaccination strategies and equitable control programmes.

In order to realise the above data collection of remote sensing, delay tolerant networks using Raspberry Pi will be deployed. Together with human contact data collection, daily based updated web information will be delivered to remote communities using an advanced web server caching mechanism.

Moreover, the project is being extended towards two directions as follows:

1) Human behavioural study: Individuals may change their behaviour for several reasons: by being ill themselves, having to care for others who are ill, or by changing their normal habits in the belief it will prevent or minimise their risk of infection. A recent study suggests that public transport usage may decline in the event of an influenza pandemic and that people may stay at home rather than go to work and risk infection. If such precautionary behaviours were to be adopted by a large number of individuals, the economic implications may be profound.

2) Efficient large-scale networked data processing: This requires heavy data processing resources when the scale of the data increases and the novel data processing for the graph data using even with single laptop has been investigated.
Exploitation Route The project will be used to develop improved mathematical models for the spread of infectious diseases, such as measles, tuberculosis and pneumococcal diseases in developing countries in Africa and Asia. These diseases are vaccine preventable and there has been a significant investment in improving vaccine coverage and introduction of new vaccines in some of the poorest countries. Funding for vaccination programmes is limited and many countries face difficult decisions to refine the effective vaccination strategies within the limited budget. We are planning to experiment our research prototype in developing countries in Africa (e.g. Malawi) to demonstrate what we can be modelled using the collected social contact data. It is our intention to develop a range of mathematical models based on contact patterns, which can be used to help guide vaccine policy development over the coming decade in Africa and other developing countries.
Sectors Agriculture, Food and Drink,Communities and Social Services/Policy,Digital/Communication/Information Technologies (including Software),Energy,Environment,Healthcare,Other

URL http://www.cl.cam.ac.uk/research/srg/netos/DDEPI/
 
Description BBC, Times, and various online websites interviewed me about the FluPhone project (e.g. http://www.bbc.co.uk/news/uk-england-cambridgeshire-13281131). The BBC website on FluPhone app 'helps track spread of infectious diseases (http://www.bbc.co.uk/news/uk-england-cambridgeshire-13281131) received the highest number of hits on that week in the world. Beneficiaries: Public audience, Government, Health agencies, etc. The research contributed to a large extent to establish the topic of 'Digital Epidemiology'. The project demonstrated the ability of measuring human proximity and building contact networks using inexpensive devices such as mobile phones, RFID tags, Raspberry Pi, and GPS. This approach especially benefits developing countries, where such information is most needed, e.g. for efficient vaccination strategies with limited resources. Additionally, the project provides internet accessibility in rural remote regions, e.g. for useful information related to health and potentially building a remote diagnosis system.
First Year Of Impact 2012
Sector Healthcare
Impact Types Cultural

 
Description GLOBAL CHALLENGES RESEARCH FUND
Amount £50,000 (GBP)
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Academic/University
Country United Kingdom of Great Britain & Northern Ireland (UK)
Start 06/2016 
End 03/2017
 
Title Data Mule in Amazon 
Description Human connectivity patterns in rural area in Brazilian Amazon 
Type Of Material Database/Collection of data 
Provided To Others? No  
Impact Rare information of human connection network in rural and remote villages in developing countries (modelling in progress). 
 
Title EEG and Eye tracking: Social influence 
Description Measurement and model of social infleunce in the brain using EEG and eye tracking. 
Type Of Material Database/Collection of data 
Provided To Others? No  
Impact rare and important research inspiration of measuring social influence. 
 
Title Mobility data 
Description Collection of human mobility data using RFID tags, Blutooth enables sensors. 
Type Of Material Database/Collection of data 
Year Produced 2013 
Provided To Others? Yes  
Impact Other researches use the data for comparison to their approaches and the research topic is promoted. 
URL http://www.crawdad.org/keyword-DTN.html
 
Description Collaboration with London School of Hygiene and Tropical Medicine 
Organisation London School of Hygiene and Tropical Medicine (LSHTM)
Country United Kingdom of Great Britain & Northern Ireland (UK) 
Sector Academic/University 
PI Contribution Collaboration with John Edmunds and Ken Eames at London School of Hygiene and Tropical Medicine, and with Jon Reed at the University of Liverpool and working on the joint funding proposal.
Start Year 2010
 
Description Distributed data processing 
Organisation Microsoft Research
Department Microsoft Research Cambridge
Country United Kingdom of Great Britain & Northern Ireland (UK) 
Sector Private 
PI Contribution Discussion on the topics in common interest and joint supervision of a PhD student, who works on related topic of the project.
Collaborator Contribution Feedback of the project work including reading papers prepared for the publication.
Impact Research feedback.
Start Year 2011
 
Description Human mobility measurement and delay tolerant communication experiments in Brazilian Amazon remote region 
Organisation Federal University of Amazonas
Country Brazil, Federative Republic of 
Sector Academic/University 
PI Contribution Joint experiments to deploy raspberry pi based stand alone networks in remote region of Brazilian Amazon
Collaborator Contribution Human mobility measurement using RFID tags and Raspberry Pi over 100 population.
Impact 2016/2017 plan of another experiment of hjuman mobility measurement and delay tolerant networks depending on funding availability.
Start Year 2015
 
Description Large-scale graph processing using external memory 
Organisation Swiss Federal Institute of Technology in Lausanne (EPFL)
Country Switzerland, Swiss Confederation 
Sector Public 
PI Contribution Joint work on SSD based external memory approach on large graph processing sepcifically on sequential access methodology.
Collaborator Contribution Random access methodology on used of SSD inlarge scale graph processing
Impact publications: K. Nilakant, V. Dalibard, A. Roy, and E. Yoneki: PrefEdge: SSD Prefetcher for Large-Scale Graph Traversal. ACM Internataional Systems and Storage Conference (SYSTOR), June, 2014. E. Yoneki and A. Roy "Scale-up Graph Processing: A Storage-centric View". ACM SIGMOD - GRADES, New York, USA, June, 2013.
Start Year 2012
 
Description Social Proximity Network Data Collection, Modelling and Analysis of TB in Tanzania 
Organisation Ifakara Health Institute
Country Tanzania, United Republic of 
Sector Charity/Non Profit 
PI Contribution We aim at contributing to our understanding of these social proximity networks, their roles in TB transmission dynamics, and methods for measuring them, where we combine carbon dioxide exchange and electronic proximity sensing to understanding social contact and quantify risk of transmission.
Collaborator Contribution Our partner reviewed existing TB transmission models to understand how data on proximity networks might be incorporated into them.
Impact Plan for funding proposal to MRC Research Council.
Start Year 2016
 
Description Storage and memory system for graph processing 
Organisation Intel Corporation
Department INTEL Research
Country United States of America 
Sector Private 
PI Contribution Analysis of graph processing specific of storage systems
Collaborator Contribution Joint papers
Impact Joint papers
Start Year 2015
 
Description Cambridge University Big Data summit 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Public/other audiences
Results and Impact At the event, where University of Cambridge initiates to give a series of lectures on recent movement on Big Data. I presented the large scale data processing perspectives.
Year(s) Of Engagement Activity 2015
URL http://www.bigdata.cam.ac.uk/resources/the-vocabulary-of-big-data/
 
Description Dagstuhl Seminar 14462 - Systems and Algorithms for Large-scale Graph Analytics 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Participants in your research and patient groups
Results and Impact Building research commnity on the topic of large scale data processing

Establishing the research integration between industry and academia

writing white paper on large sclae graph processing

organisation of workshops
Year(s) Of Engagement Activity 2014
URL http://www.dagstuhl.de/no_cache/en/program/calendar/semhp/?semnr=14462
 
Description FluPhone Project 
Form Of Engagement Activity A magazine, newsletter or online publication
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Public/other audiences
Results and Impact http://www.cl.cam.ac.uk/research/srg/netos/fluphone2/.
Year(s) Of Engagement Activity 2011
URL https://www.fluphone.org.
 
Description Keynote talk at Complex Network Conference 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Future collaboration in Complex Network and MachineLearning Research topics.
Year(s) Of Engagement Activity 2016
 
Description Seminar at KTH, Sweden 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Giving a presentation to the seminar series in KTH, stockholm, Sweden.
Year(s) Of Engagement Activity 2014
 
Description Seminar at University of Helsinki 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Discussion summary on future mobile Internet
Year(s) Of Engagement Activity 2016
 
Description Seminar to General Public - Pint of Science 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Public/other audiences
Results and Impact connect science and general public - good conversation after the talk.
Year(s) Of Engagement Activity 2016
 
Description iSocial: Marie Curie Initial Training Networks: Keynote lecture 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Schools
Results and Impact The wrokshop is organised for mainly graduate students working on the topics of Online Social networks: Data Management and Privacy/Security Issues. The expert researchers from those reserach topics gave a series of lectures, and there was a good level of research discussion during the lecture time as well as social occasion. Many future collaborations were discussed.
Year(s) Of Engagement Activity 2015
URL http://linc.ucy.ac.cy/isocial/