Meet MILTON, AZ's AI that can predict 1,000+ diseases

A machine-learning tool developed by AstraZeneca and trained on UK Biobank health record data could predict over 1,000 diseases, before diagnosis, according to a paper in Nature Genetics.

Called MILTON – short for MachIne Learning with phenoType associatiONs – the AI can predict whether individuals are more likely to have and be diagnosed with certain diseases based on 67 biomarkers that are routinely collected during clinical practice. That includes blood biochemistry, blood counts, respiratory function scores, blood pressure variables, and other measures like age, sex, body size, and fasting time.

In the study, MILTON was used to analyse 3,200 diseases and achieved a high prediction score for 1,091 of them. It can be applied to any biobank, irrespective of genomic ancestry, and will be further developed by adding additional data such as proteomics collections.

The researchers behind the study say that MILTON has the potential to accelerate the discovery of new drug targets and biomarkers, paving the way for the development of more effective and targeted treatments, as well as allowing early disease detection.

In conventional case-control studies, hospital billing codes and self-reported data often classify participants into cases and controls, which can be incomplete. However, MILTON enables users to identify individuals who may have been incorrectly classified as controls. 

This AI-augmented reclassification significantly enhanced the statistical power for genetic discovery, expanding the scope and accuracy of gene discovery for hundreds of diseases, said the researchers.

According to lead author Slavé Petrovski, head of the Centre for Genomics Research at AZ, MILTON is a significant advance on the predictive tools currently used and can outperform gene-based risk scoring systems.

"Our research demonstrates MILTON's capabilities and how it is able to identify disease risk cases in large biobank datasets, which in the future, could enable us to detect illnesses earlier and at more treatable stages," said Petrovski. 

"Improving our ability to detect illnesses earlier and at more treatable stages is critical for early interventions in clinical care."

Other experts have sounded a note of caution, however. Professor Tim Frayling, professor of human genetics at the University of Geneva, applauded the thoroughness of the study, but said care needs to be taken when talking about predicting disease when "we really mean 'we can give you a slightly better idea of your chances of developing a disease, but there are still many unknown factors'."

Professor Dusko Ilic, a stem cell specialist at King's College London (KCL), said MILTON represents a "significant step forward in the field of predictive medicine", but voiced some concerns about its ethical use.

"The powerful predictive abilities of this tool could, if unregulated, be misused by health insurance companies or employers to assess individuals without their knowledge or consent," he suggested.

"This could lead to discrimination and a breach of privacy, [so] strict guidelines and oversight will be critical in ensuring that the benefits of MILTON are realised in an ethical and responsible manner."