New programme to evaluate AI systems for antimicrobial resistance
The Fleming Initiative is delighted to announce a new three-year programme, supported by Google DeepMind, to develop the methods and standards needed to evaluate artificial intelligence (AI) systems for antimicrobial resistance (AMR).
AI systems are already transforming scientific research. One example is Google DeepMind's AlphaFold, which can accurately predict the three-dimensional structures of proteins – a breakthrough recognised with the 2024 Nobel Prize in Chemistry. Its predictions are now freely available to researchers around the world, supporting discoveries across biology and medicine. Progress like this relies not only on advances in AI, but also on robust ways to evaluate and compare how well different AI systems perform.
The same need now exists in AMR. AI is increasingly being used to identify drug-resistant pathogens, discover new antibiotics, track the spread of resistance and support clinical decision-making. However, there are currently no widely accepted standards for assessing whether different AI systems are accurate, reliable and suitable for these tasks. This currently makes it difficult for researchers, healthcare providers and policymakers to compare approaches, understand their strengths and limitations, or decide when they are ready for use.
Building on the Fleming Initiative and Google DeepMind report, Harnessing Artificial Intelligence to Tackle Antimicrobial Resistance, and the Google DeepMind Academic Fellowship in AI and AMR – whose inaugural recipient has now been appointed as Assistant Professor in Biomedical Electronics at the Department of Electrical and Electronic Engineering, Imperial College London – the programme will develop evaluation frameworks and benchmarking approaches for priority AMR applications. These will define how AI systems should be tested and what evidence is needed to demonstrate that they perform reliably across different populations, healthcare settings and datasets.
The programme will also work with the global AMR community to encourage data to be collected, curated and reported in ways that support robust evaluation. By establishing common standards, the programme aims to make it easier for researchers from around the world to compare AI approaches, build on one another's work and develop systems that can be used with greater confidence in research and healthcare.
Professor Alison Holmes, Director of the Fleming Initiative says:
“Artificial intelligence has enormous potential to help us tackle antimicrobial resistance. But to make a meaningful difference, researchers, healthcare professionals and policymakers need to be assured that AI systems are accurate, reliable and appropriately representative of the populations and settings where they will be used. Thanks to support from Google DeepMind, we are helping to build the foundations needed to evaluate AI consistently, enabling promising innovations to be adopted with greater confidence and deliver real benefits for research, healthcare and public health.”
Agata Laydon, Science Lead, Google DeepMind Impact Accelerator, says:
“Safeguarding global health security from antimicrobial resistance requires continuous innovation. AI systems and approaches can fundamentally change how we understand and counter this threat, but responsible model development and deployment require robust, objective evaluation. This is key to translating potential into real-world impact. We are delighted to support the Fleming Initiative as they build this critical infrastructure, providing the foundation the global research community needs to drive scientifically grounded progress.”
