TL;DR

IMU Biosciences, a London company that combines immune profiling with AI analysis, is working with UK Biobank on an 80,000-sample study of the immune cells in participants’ blood, R&D World reports. The study, announced on 19 October, will link those measurements to participants’ genetic data and health records to test which immune differences go with which illnesses. A UK Sovereign AI grant is paying for the first 10,000 samples.

How the study works

Half the samples are fresh and half come from frozen stores, and profiling got under way in September. The frozen blood is the clever part: because it was collected earlier, researchers can compare a person’s earlier immune profile with illnesses they went on to develop, which could reveal patterns that come before disease.

IMU plans to chart how the mix of immune cells shifts with a person’s age, sex and medical history. The result would act as a reference, letting researchers judge whether a given pattern is normal for a particular group. The resource already stores measurements of circulating proteins and metabolites, plus blood-cell counts, so fine-grained immune profiles would give researchers a further dimension to work with.

Waters is providing flow cytometry and single-cell multiomics equipment, namely its BD Rhapsody and BD FACSDiscover systems, and Ultima Genomics, based in Fremont, California, is supplying sequencing on its UG200. IMU handles the computational analysis.

Why the AI framing matters

The Sovereign AI-funded first 10,000 samples will also yield transcriptomic data at single-cell level, showing which genes are active in each cell. According to IMU, a single platform and protocol will be applied across the whole group, so that AI models have consistent measurements to learn from, each tied to a participant’s clinical, genetic, lifestyle and imaging records.

Looking forward

The partners plan to return the data to UK Biobank, where approved researchers will be able to use it, and IMU’s longer-term aim is to profile all 500,000 participants. On 15 October we reported a separate Sovereign AI award, to Augur Security. In our view, this is one of the more defensible uses of sovereign AI money: it builds a national dataset that UK researchers and start-ups can train models on, rather than funding a single product. The value will depend on how quickly and cheaply researchers can get access.