Sovereign AI supports OpenImmune to build an open dataset for predicting immune reactions to medicines
Through our Strategic Assets Programme, Sovereign AI is supporting OpenImmune with £5 million to build an open dataset that could help AI models better predict how the human immune system responds to biologic medicines. OpenImmune is a UK-led, not-for-profit Focused Research Organisation.
The project will address a significant challenge in drug development: predicting whether a patient's immune system will react against a treatment. By bringing together experimental data and computational modelling, OpenImmune aims to create a resource that helps researchers and drug developers develop better predictive models and design safer, more effective medicines.
The resulting dataset, models, benchmarks and code will be hosted in the UK and made openly available to companies and researchers across the ecosystem, in the UK and worldwide, with no preferential access for anyone, giving them the opportunity to build on the work.
AI Minister, Kanishka Narayan, said:
“Medical breakthroughs are slow, painstaking work, and the wait for new therapies is simply too long for patients living with diseases and the suffering they cause. AI could change that. It can help researchers pull breakthroughs out of vast amounts of data, faster than has ever been possible before.
“But those leaps forward are only possible if we have the data AI needs. This work could make a big difference to how we understand immune reactions, helping us develop safer and more effective treatments. We’re supporting OpenImmune to build that body of knowledge, so that health researchers can use it to unlock breakthroughs from which we’ll all ultimately benefit.”
Building the data behind better AI models
Biologic drugs are medicines produced using living systems, often based on complex biological molecules such as antibodies. They are increasingly important in treating a wide range of diseases, but predicting how the immune system will respond to them remains a major challenge.
An immune response against a biologic drug can reduce its effectiveness and, in some cases, cause harmful side effects. These reactions can also mean patients need to switch to alternative treatments, which can take time. For developers, an unexpected immune response can end a promising medicine late in development, after years of investment.
Despite the importance of the problem, the data needed to train and evaluate predictive AI models has never been generated at the scale, or across the range of human genetic diversity, that prediction needs.
OpenImmune will tackle this gap by linking the molecular sequences of biologic drugs with experimental measurements of the immune responses they generate. Combining these measurements with computational modelling will help researchers investigate patterns in immunogenicity and develop models that can better predict potential immune reactions.
By making the resulting resources openly available, the project aims to give AI developers access to data at a scale that no single organisation could generate alone.
An asset for the wider AI ecosystem
The value of the project extends beyond OpenImmune itself. Its outputs are intended to support a wider community of AI companies and researchers applying AI to drug discovery.
Techbio and life sciences companies including Boltz, Illumina, IMU Biosciences, Latent Labs, Antiverse and Serova Bio are engaged with the project as observers, advising on the science with no financial commitment and no preferential access.
The programme is delivered by specialist partners contracted by the OpenImmune Foundation: Arctoris, DaltonTx, RepertoireBio, and two University of Oxford groups, the Oxford Protein Informatics Group and the Fernandes Laboratory.
Access to the dataset and associated tools will allow researchers to test new approaches and build on a shared evidence base. The ambition is to create a resource that becomes more valuable as more teams use it to develop and improve their models.
The project will also help establish valuable research capabilities in the UK, strengthening the foundations on which British AI and life sciences companies can innovate and compete globally.
Tom Fleming, Founder and Chief Executive of OpenImmune, said:
“Immunogenicity is one of the largest unsolved problems in medicine, and it has stayed unsolved because the data needed to predict it has never been generated at the scale required. This award lets us start building it, openly, in the UK, for everyone. It is the first part of the funding for the programme, and the part that makes the rest possible; we are now seeking philanthropic partners to complete it.”
Supporting the foundations of British AI with Strategic Assets
Sovereign AI's Strategic Assets Programme supports projects that create valuable datasets, infrastructure and capabilities for the wider UK AI ecosystem.
Some of the most important resources for AI development are difficult for individual companies to build alone. They can require substantial upfront investment and collaboration across organisations, even where the benefits could extend across an entire field.
Through the programme, Sovereign AI provides non-equity funding to help bring these projects forward and make their outputs available to a wider community of innovators. We're pleased to be working with OpenImmune to build this important new resource for the UK's AI and life sciences ecosystem. We look forward to seeing what the team delivers and how researchers and companies across the field put it to use.
About OpenImmune
OpenImmune is a UK-led, not-for-profit Focused Research Organisation building the world's first open, foundational dataset and AI model for predicting the immunogenicity of biologic medicines. The OpenImmune Foundation holds every result and releases it openly, with no paywall and no preferential access. Find out more at OpenImmune.ai.