Biosurveillance via ImmunOlogical, Genomic and Epidemiological Models

Risk modelling integrating mobility and biological data, assessing threats from variants of concern and driving policy through data synthesis.

Principal Investigator: Madhav Marathe, Ph.D.

Division Director and Distinguished Professor in Biocomplexity, Biocomplexity Institute & Initiative (BII)

Professor of Computer Science, School of Engineering and Applied Science

University of Virginia

Genetic drift and selection pressure in the SARS-Cov-2 genome (specifically the spike protein) have increased transmissibility and virulence, fueling a resurgence in COVID-19 globally. The continuing interplay among viral evolution, human behavior, and public policies will determine the future course of this pandemic.  

Problem: The spread of SARS-CoV-2 has highlighted the need for enhancing integrated bio-surveillance capabilities: sequencing viral samples across space and time, combining this information to assess the distribution and impact of the viral strains, predicting which mutants are likely to become dominant over time, and identifying the best measures to control them.  Such an integrated capability is needed for effective pandemic planning, response, and mitigation strategies. Interventions should reflect geographic and sociological scales and must be adaptive.  Currently such an integrated capability does not exist -- current approaches miss critical early warning signals and produce piecemeal solutions which use scarce resources inefficiently.

Hypothesis: Data including genomic, immunological, climatic, ecological, and sociocultural can inform multiscale/multispecies models and simulations.  Stochastic models can integrate epidemiological, immunological and viral evolutionary data, building a picture of the spatio-temporal-demographic dynamics of viral sequences. Such a system is needed to inform the deployment of a sensor network that promotes human health and safety. 

End users and Stakeholders: Our solution will serve epidemiologists, bioinformaticians and policymakers by providing them with a comprehensive and intuitive online tool suite that can characterize emerging viral threats. It will enable threat assessment with the most accurate models to inform public policy including threat assessment for dominant strains, prediction of efficacy of ’lockdown interventions’, social restrictions and vaccination strategies.   The proposed solution will simultaneously model observed patterns and threats and also help answer specific questions regarding the transmission dynamics of variants of concern in the context of extensive vaccination (US, UK) or less (India) under very different population demographics, growth of infection and transmission patterns.

What needs does it address: Our solutions and tools will take a step toward understanding the complex interactions among human mobility, social distancing, vaccine hesitancy, viral evolution and testing, and vaccine design and deployment. We acknowledge the fact that a full solution will take much more time and effort.  As the pandemic continues to evolve, this understanding will become increasingly important for informing targeted intervention design and intelligently guiding surveillance and vaccine deployment.

BIOGEM will enable significant reductions in the global burden of infectious diseases. As noted by Fineberg and Wilson (Science 2009), few situations illustrate the salience of the chain “theory to science to policy'' more dramatically than a pandemic. The need for comprehensive biosurveillance is also discussed in various reports (e.g. Bedford et al. Nature 2019). We expect the proposed system to provide the following public goods:

• An integrated bio-surveillance system that can better detect variants of concerns leading to fewer cases and ultimately hospitalizations and deaths. 

• A more efficient and targeted bio-surveillance system that can potentially scale up and be more cost effective.

• We have established that repeated doses of existing vaccines will expand the immune repertoire and neutralisation breadth with respect to SARS-Cov-2. Understanding the threat level posed by new variants of concern, and the likely effects differing vaccination strategies have in terms of reactogenicity and minimising future epidemic spikes, will allow targeted public health interventions with minimal disruption. In addition, we have identified antibody decay curves, in-home transmission dynamics and will continue to add the available biological datasets as needed to ensure efficient quantification of biological responses and improve accuracy of computational modelling where available. 

Team members have been informing policy at local, regional and national level in their respective nations (USA, UK, India) including :

• Supply of biological observations and briefing documents to governmental disaster management groups.

• Development of regional and national models of disease dynamics feeding into governmental policy.

• Development of Google community and mobility reports. Globally, these have become the reference data for effectiveness of lockdowns and social restrictions

Linking these approaches will provide tangible impact by allowing direct feed-in of biological observations, epidemiological data, mobility and activity data and viral surveillance data into a single approach informed by all appropriate sensors into a bespoke modelling engine with a single simple online interface will allow policy makers from the entire health and social care delivery setting. We envisage real time use ranging from health providers, e.g. family doctors, hospital boards to social and governmental national policy makers.

We will start with a small focused area, likely in the US. Experience working with the Virginia Department of Health throughout the pandemic makes this state a logical possible starting point.  Within three months we will grow the area of focus to the UK.  The overall system architecture will be designed to scale to one site and then to all the sites over one year.  India in particular will allow a massive increase in scale.  The population of the state where IISc is located has by itself a population of seventy million. 

BII has significant experience modeling the population of the US with county-level granularity.  While the system can (with the proper data) be focused on smaller segments, the greatest impact for our solution seems to lie in country-scale modelling with county (or similar administrative unit) detail.  

Further expansion is possible into Brazil and France in the later years of the project.  This will be dependent upon project success as well as the correct partnerships, availability of data, etc. 

Our primary impact goal is to inform targeted surveillance and mitigation policy at regional and national levels. Because many trade-offs are involved in these decisions, measuring the impact of any one body of evidence is not straightforward. Historically, we have found the best metric to be simply adoption and continued use of the technology by policy makers. This, in turn, requires building confidence in the situation assessments and forecasts provided by the tool. We will use the usual kinds of retrospective analysis and sensitivity testing to analyze the quality of our situation assessments in aggregate form as well as the individual components of the framework, but by definition, we cannot evaluate the performance of the tool in the kinds of counterfactual scenarios that must be considered in finding optimal strategies. We can, however, within the framework, demonstrate the value of information as embodied in the tool through retrospective analyses: How much earlier would a novel threat have been detectable with the surveillance strategies suggested by this tool? Could we have controlled viral evolution better with intervention strategies that generated different spatio-temporal-demographic selection pressures? We will also validate forecasts prospectively against viral and epidemiological dynamics during the course of the pandemic.

Possible barriers to BIOGEM success:

• Availability of data in a timely manner. We have set up partnerships with various organizations and expect to overcome this challenge.  Trinity may be able to assist in this area as well.  

• Sharing data across multiple groups. Again we have just completed an agreement with University of Nottingham and also completed an agreement with IISc. This gives us confidence that we can continue to develop multi-country data sharing agreements. 

• Ability of government agencies to work closely with us to change their sampling and sensing protocols. In the past, certain operational constraints made this challenging but we believe that the COVID-19 pandemic has brought to the fore the need to be flexible.

• Teams located in multiple time zones. We have found ways around this as a part of our other ongoing projects.  BII has extensive experience coordinating large research programs across many institutions and countries.

• Continued remote work makes bringing the team together a bit more challenging. But technologies developed in the last few years will alleviate some of the issues.

Biocomplexity Institute and Initiative, University of Virginia, USA

University of Nottingham, UK

Indian Institute of Science/Indian Statistical Institute, India

Princeton University, USA

In addition, we have partnered with Google research and Strand Life Sciences in India as well as the Virginia Department of Health in the US.

We are applying to the Trinity challenge for following reasons:

1. The challenge provides significant resources that can be leveraged with ongoing work by the team, providing a unique opportunity to develop the next generation global biosurveillance system.

2. The funding is flexible and allows us to form international collaborations to work on a problem that by its nature is global. Other funding sources have largely funded each of the teams individually. 

3. The recent upsurge in COVID-19 cases in the US, UK, India and other parts of the world, new variants and their potential impact and rollout of vaccines along with their positive impact but also with the potential to create new evolutionary pressures, make this a timely problem that we felt was appropriate.

4. The unique ability to create a completely open system and generate transparent and freely available data sets. Often due to various constraints this is somewhat challenging for certain sponsors.

5. Prestige: The Trinity Challenge is prestigious.  We think funding from this competition will open doors for future funding sources as well.

In addition to the list of partners listed above, we would like to partner with pharmaceutical companies and organizations that are involved in sampling and sequencing.