Their work, published in Nature Medicine, could be a key-stepping stone toward unifying the understanding of this disease by providing a data-driven examination of MS disease evolution. The results may also have practical implications for the management of MS and for the discovery of new therapies.
MS affects around 2.9 million people around the world. There is no cure for the condition; available treatments can reduce the risk of neurological symptoms and hinder the disease from worsening. However, over time, most patients progress and accumulate disability. The treatment of progression remains an urgent unmet medical need.
Disease classification impacts people living with MS, as it defines who is eligible for available medications. It also influences the discovery of new therapies, as it defines the populations that must be studied in clinical trials. However, the traditional way that MS has been divided into distinct subgroups has not accurately reflected the underlying biology of the disease. Increasingly complex and divergent views and sub-classifications have been proposed but there remains a need for a unified approach to classification that meaningfully reflects disease pathology.
In their study, the multidisciplinary team developed a bespoke AI model and interpreted the data from over 8,000 people with MS who were followed for up to 15 years to examine disease evolution over time. The work aimed to either confirm the traditional subtyping of MS, or to propose a new data-driven classification based on the pattern of progression observed in these patients. The results were validated in independent clinical trial and real-world data from another 4,000 people with MS.
The team discovered that MS is likely one disease, characterised by a gradual shift in disease biology from early, mild and evolving states to more advanced states. Dynamic episodes of inflammation - regardless of whether these are accompanied by symptoms - drive the accumulation of damage to the brain and the physical and cognitive worsening of health over time.
Asymptomatic or undetected disease activity has been identified as a major driver of disease progression across the full spectrum of MS. To protect the brain and preserve function, treatment with efficacious therapies is essential including during these early stages.
The team also found that patients in advanced states of MS, who are progressing in their disease, can be considered as a single group for purposes of drug development, as the mechanisms leading to worsening disease seem to be the same. This understanding could inform how clinical trials are conducted and the precision with which medications are prescribed.
Dr Habib Ganjgahi, Senior Researcher in Statistics: Statistical Machine Learning and Image Analysis at the Big Data Institute, said: ‘We have developed an AI method, grounded in probabilistic machine learning, that offers a unique, unified view of multiple sclerosis. By integrating diverse data over time, our approach helps reclassify the disease in a way that could guide prognosis and accelerate the development of future treatments. Importantly, this framework is not limited to MS—it can be applied to other conditions with longitudinal data, opening the door to broader insights across many diseases.’
‘This study looked at over 8,000 people with MS to better understand how the disease changes over time,’ said Dr Dieter Haering, Executive Director of Biostatistics at Novartis. 'The results show that MS is more like a continuous journey than separate types, which could help in creating better treatments for patients at every stage of the disease.’
Read the full paper on the Nature Medicine website: https://www.nature.com/articles/s41591-025-03901-6