microscopic image of coronavirus COVID-19

Studies in Pandemic Preparedness

These collaborative projects with academia span topics including infection prevention and control, treatment and diagnostics, mental health, and return to work.

Infection Prevention and Control Projects

  • University of Oxford: Deirdre Hollingsworth
    MIT: Chris Rackauckas
    Alan Turing Institute: Andrew Duncan (Imperial College London)
    Swansea University: Michael Gravenor, Biagio Lucini
    University of Cambridge: Ronojoy Adhikari
    University of Warwick: Sebastian Vollmer (Alan Turing Institute)
    Royal Society Rapid Assistance in Modelling the Pandemic (RAMP): Graeme Ackland (University of Edinburgh), Kostas Kavoussanakis (University of Edinburgh), Mike Cates (University of Cambridge)
    Microsoft: Simon Frost, Tim Carroll (Azure HPC), Rolf Harms (Cloud & AI), Grace Huynh, Oege de Moor (GitHub), Aqeel Siddiqui (GitHub), Kenji Takeda, Miah Wander, Jenny Ye (Cloud & AI)

    Mathematical modeling of infectious disease transmission is an important tool in forecasting future trends of pandemics, such as COVID-19, and in evaluating ‘what if’ projections of different interventions. Yet, different models tend to give different results. To draw robust conclusions from modeling, it is important to consider multiple models, which can be facilitated by expanding the modeling community, as models tend to reflect the people that develop them, and by bringing models and model developers together to compare and contrast models. The aims of this project are: to expand and empower the next generation of models and modelers; to act as an incubator for a platform for comparisons of infectious disease models, both in terms of computers and people; and hence enable swift and more robust policy decisions for current and future pandemics.

  • University of Washington: Shwetak Patel, Luis Ceze
    Microsoft: Jonathan Lester, Bichlien Nguyen, Karin Strauss, Mike Reddy, Asta Roseway, Mike Sinclair

    The Molecular Information Systems Lab (MISL) and the UbiComp Lab at the University of Washington in collaboration with Microsoft Research are exploring hybrid molecular-electronic systems, with a particular focus on DNA computing as a hybrid approach for detecting COVID-19 in the environment. Detection of the virus has been demonstrated in city wastewater samples. As such, it has been proposed that environmental sampling locations, such as wastewater or air filtration systems, could be informative indicators for continuous monitoring to help cities stay ahead of community-wide COVID-19 spread. Towards that end, the objective of this research is to develop environmental sampling and pathogen detection techniques for air filtration systems on public transit networks, such as buses and subway systems. These techniques include safe sample extraction, qPCR, DNA sequencing, real-time nanopore protocols, and molecular computation. This offers the potential to detect the virus in buses, airliners and buildings, enabling new avenues for viral spread monitoring for a safer and better-informed pandemic response.

  • University of Washington: Shwetak Patel, Luis Ceze
    Microsoft: Bichlien Nguyen, Jonathan Lester, Karin Strauss, Mike Reddy, Asta Roseway, Mike Sinclair

    There is currently a shortage of medical-grade face masks worldwide, and that shortage is likely to continue for the duration of the pandemic. As a result, the CDC is providing guidance on how people can create homemade face masks to mitigate the spread of COVID-19. There have been many recent efforts to evaluate the filtration quality of various everyday materials. The objective of this research is to leverage the sensors embedded on commodity smartphones to assess the filtration capability and breathability of homemade face masks. This effort will culminate in a mobile app that can be used to help end-users create better homemade masks, which would significantly limit the risk of virus transmission as social distancing measures begin to relax worldwide.

  • Johns Hopkins University: Anton Dahbura
    Royal Society Rapid Assistance in Modelling the Pandemic (RAMP) & University of Edinburgh: Graeme Ackland, Steven Carlysle-Davies, Kostas Kavoussanakis
    University of Exeter: Peter Challenor, Michael Dunne
    University of Oxford: Deirdre Hollingsworth, Emma Davis, Jasmina Panovska Griffiths, Andreia Vasconcelos
    University of Washington: David Smith, Sean Wu
    The Wilson Centre: Alex Long
    Microsoft: Xiaoji Chen, Simon Frost, Eric Horvitz, Kenji Takeda, Jessica Young

    Computer-based epidemiological models are important tools to help policymakers make important public health decisions. There are, however, challenges in both determining model uncertainty and communicating this uncertainty effectively to policymakers and the public. This project brings together a multi-disciplinary team of academic and Microsoft researchers to develop, demonstrate, and deploy novel methods to better quantify and explain uncertainty in epidemiological models. The goal is to help improve public health decision-making and communication at national, regional, and local levels to effectively respond to COVID-19, future pandemics, and vector-borne epidemics via Microsoft Premonition.

Treatment and Diagnostics Projects

  • University College London Hospitals NHS Foundation Trust (UCLH) and University College London (UCL): Joseph Jacob, Sam Janes, Daniel Alexander, Geoff Parker, Jerry Brown, Arjun Nair, Paul Taylor, David Hawkes, Marc Lipman, Joanna Porter, John Hurst, Nick McNally, Bryan Williams
    Microsoft: Javier Alvarez-Valle, Usman Munir, Melissa Bristow, Melanie Bernhardt, Anton Schwaighofer, Jay Nanavati, David Carter, Ozan Oktay, Shruthi Bannur, Hannah Murfet, Kenji Takeda, Aditya Nori

    Microsoft Research Project InnerEye team in Cambridge (UK) is supporting UCLH and UCL to help identify vulnerable patients who are not currently covered by “shielding” guidelines. Additionally, UCLH and UCL will help identify patients for whom shielding may not be necessary to avoid economic hardship imposed by unnecessary shielding in future Covid-19 outbreaks. UCLH and UCL will evaluate pre-Covid-19 imaging and clinical data to predict endpoints and speed up development of imaging biomarkers. They will look at cardiac computed tomography (CT) and use quantitative techniques to identify pre-Covid-19 cardiac imaging metrics that link to the likelihood of severe Covid-19 infection and cardiac events. The InnerEye team will support UCLH and UCL, who will be developing models for the Prognosis of COVID-19 using CT imaging. This work is supported by the National Institute for Health Research Biomedical Research Centre at UCLH.

  • University Hospitals Birmingham NHS Foundation Trust: Shazad Ashraf, George Gkoutos, Andrew Beggs, Kal Natarajan, Elizabeth Sapey, Alastair Denniston, Sharan Wadhwani
    Microsoft: Javier Alvarez-ValleMelissa Bristow, Melanie Bernhardt, Anton Schwaighofer, Jay Nanavati, David Carter, Ozan Oktay, Shruthi Bannur, Junaid Bajwa, Usman Munir, Hannah Murfet, Kenji Takeda, Aditya Nori

    Microsoft Research’s Project InnerEye team in Cambridge (UK) is working with University Hospitals Birmingham NHS Foundation Trust to develop deep learning models to analyze anonymized chest X-Rays and chest computed tomography (CT) scans to assist clinicians in determining disease severity, aid decision making, and improve our understanding of the disease. The aim is to improve the objective determination of disease severity by classifying and quantifying lesions in the lungs. This could help provide additional information for making a prognosis, assisting in the management of both hospitalized patients and their longer-term health needs. By quantifying disease progression, the model may aid hospitals in making decisions about resource deployment. It will also help to build our knowledge of the disease.

  • Stanford University: Allison Koenecke, Ruoxuan Xiong, Susan Athey
    Johns Hopkins University: Mike Powell, Joshua T. Vogelstein
    Microsoft: Weiwei Yang, Emre Kiciman, Chris White

    Prospective randomized clinical trials are the most reliable way of ascertaining the causal effect of a treatment on patient outcomes. However, trial prioritization for both institutions and individuals remains a complex problem due to limited numbers of highly heterogeneous patients. This project will conduct federated retrospective analyses designed to assess the benefit of off-label drug use by pooling multiple disparate databases, to help prioritize and guide subsequent initiation and recruitment of randomized clinical trials. This will include evaluating the impact of the target drugs on patient outcomes from diseases similar to COVID-19, such as pneumonia or acute respiratory distress, generating artificial datasets using generative adversarial networks to asses performance of methods when ‘ground truth’ is known, applying the best methods to analyze the effect of the target drugs on the outcomes of COVID-19 patients across hospital systems, and using the results to evaluate the potential of these drugs and suggest guidelines for clinical trials.

  • University of Washington: Jesse Erasmus, Deborah Fuller
    Tufts University: Charles Shoemaker
    InBios, Inc: Syamal Raychaudhuri
    Microsoft: Grace Huynh

    During an emerging infectious disease outbreak, rapid deployment of an effective and highly specific therapeutic can dramatically alter the course of the epidemic. One gold standard approach is to use monoclonal antibody (mAb) therapy. Traditional approaches can take months or years to execute. The protein production and purification process can take months to develop and optimize and the result may not be a true representation of the target pathogen. In this project we propose a method to accelerate the production of high-quality mAb therapeutics by eliminating the need to produce recombinant protein antigens. To screen and identify candidate mAbs without a protein antigen, we propose here a sequence-based bioinformatic and machine-learning approach to rapidly identify candidate mAbs. If successful, this method will significantly accelerate the discovery and delivery of high quality mAb therapeutics specifically for treatment of SARS-CoV-2 infection, and establish a platform that can be applicable in future outbreaks of infectious diseases. In addition, our bioinformatic approach can easily be extended to other forms of monoclonal antibody therapy, including treatments for cancer, autoimmune diseases, and neurological conditions such as Alzheimer’s disease.

Mental Health Implications and Return to Work Projects

  • Rice University: Akane Sano, Fred Oswald
    Baylor College of Medicine: Nidal Moukaddam
    Microsoft: Mary Czerwinski, Daniel McDuff, Shamsi Iqbal

    This project aims to develop an intelligent agent for a wide range of workers during the COVID-19 pandemic, to manage both their work- and life-related activities. We will leverage intelligent agents, ubiquitous and affective computing, and machine learning that analyze users’ behavior and emotion, to maintain or even improve their productivity and wellbeing in comparison with work and life outside of the pandemic. We will ask the following research questions: (1) how has the pandemic changed people’s work and lifestyle patterns and how have people managed to maintain their productivity and wellbeing? (2) can a personalized conversational agent help users manage their work/personal tasks and schedules better to improve their productivity and wellbeing during the pandemic and the transition phase of return to quasi-normal? (3) can the agent identify regular and irregular behavioral patterns related to productivity and wellbeing and usefully intervene on the user’s behalf?

  • Johns Hopkins University: Avanti Athreya, Youngser Park, Carey Priebe
    Microsoft: Carolyn Buractaon, Nick Caurvina, Darren Edge, Jonathan Larson, Chris White

    With more than 30 million claims to unemployment this year due to the pandemic, organizations are desperate to figure out how to get people back to work. This project will use network machine learning on enterprise communication and collaboration data to evaluate and detect network resilience in a post-pandemic world. This research is aimed at providing better guidance as many companies transition from “work-from-home” back to their workplaces. For example, given a large tech company, can we determine which groups’ productivity would benefit the most from returning to the office first vs which groups might be more resilient to working from home? This research will contribute network machine learning that will better inform companies on how to operate.