Accelerating infectious disease modelling using general purpose GPUs

The MRC Centre for Global Infectious Disease Analysis (MRC-GIDA) is a world-leader in infectious disease modelling and training, collaborating closely with public and global health agencies such as the WHO and governments worldwide. With over 200 researchers, we have a unique capacity to respond to emerging threats with real-time analysis and predictive modelling, providing evidence-based input for urgent policy decisions on endemic and emerging diseases such as COVID-19, Ebola, mpox, and influenza. This proposal will extend our research capabilities with high-performance general-purpose Graphical Processing Units (GPUs). These resources will enable transformative advances in infectious disease modelling leading to faster public health responses to the increasing threats of infectious diseases in the UK and worldwide. Currently, our high-performance computing (HPC) resources are CPU-based. There is an urgent need for infrastructure capable of supporting advanced machine learning techniques - methods crucial for the next generation of infectious disease research. The proposed GPU technology will allow us to develop methods that can significantly improve our response times in meeting these demands. The primary objective of this project is to establish GPU infrastructure that our diverse research community will use to transform the computational approaches relied on when responding in real-time to public health emergencies. These will reduce turnaround times for critical analyses, enhance pandemic preparedness, and support policy-making with high-resolution insights into health inequalities and complex disease dynamics. Additionally, the infrastructure will foster cross-institutional collaborations with other MRC centres in Imperial, providing a shared resource for initiatives focussing on modelling the health effects of environmental hazards, air pollution and climate change. Importantly, transitioning from CPUs to GPUs will enhance environmental sustainability by enabling faster, energy-efficient computation, reducing the carbon footprint of our research. Centre researchers will leverage the GPU infrastructure to drive advancements across a range of applications. These include machine learning models capable of emulating mechanistic infectious disease models, significantly reducing computational times and allowing real-time modelling of interventions during emerging outbreaks. These emulators can be orders of magnitude faster, enabling more detailed representation of transmission dynamics and key socioeconomic determinants of health inequity. More nuanced intervention strategies can also be designed, ensuring targeted, equitable interventions for future pandemics and endemic diseases disproportionately affecting marginalised groups, such as TB and sexually transmitted infections. In genomic epidemiology, GPUs will also facilitate evolutionary analyses, vaccine design and development of novel algorithms to manage the growing volume of genetic sequence data underpinning surveillance systems. Lastly, GPUs will advance climate and health modelling by examining the effects of environmental changes on vector-borne diseases such as malaria and dengue. The benefits will reach beyond MRC-GIDA. National and international impacts will be delivered through existing partnerships with academic institutions and agencies such as UKHSA, WHO, The Global Fund, Gavi, and ministries of health across the world. With the ability to deploy lightweight GPU-enabled models, we can also support more requests during health crises and democratise access to complex infectious disease modelling, especially in lower-resource settings. The software tools will be developed as open-source packages for researchers worldwide, and we will support their dissemination through short courses, bespoke training at international conferences, and hackathons. This will ensure advanced computational tools can be used globally, bridging technology gaps, and continuing to position the UK at the forefront of global infectious disease modelling and response.