Novel Tool Predicts Most Effective Multiple Sclerosis Medication for Patients

Symptoms can include difficulty walking, loss of muscle strength, memory problems, and mood disturbances. Although natalizumab is widely used to reduce relapses and slow progression, around 35% of patients fail to fully respond, risking relapse within two years and experiencing adverse effects. Researchers have now identified a new method to predict treatment response before therapy begins.

A collaboration between Brazilian researchers from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP, São Paulo, Brazil) and French institutions has led to the development of a predictive tool that combines advanced cell imaging and machine learning. The drug natalizumab, a monoclonal antibody, prevents immune cells from entering the brain and triggering inflammation by blocking VLA-4 from binding to VCAM-1.

Researchers observed that natalizumab alters immune cells such as CD8+ T cells, making them more rounded through actin remodeling, a protein critical for cell structure and movement. Using high-content imaging (HCI), scientists analyzed more than 400 morphological features of T cells exposed to natalizumab in vitro. Samples were taken from untreated MS patients in France and tested on VCAM-1-coated plates under immune stimulation. Machine learning algorithms processed over one million combinations, with 130 features providing relevant information for response prediction.

The findings, published in Nature Communications, show that the tool achieved 92% accuracy in the discovery cohort and 88% in the validation cohort in predicting patient response to treatment. CD8+ T cells were identified as a critical subpopulation, with non-responders displaying resistant actin remodeling, reduced polarity loss, and enhanced migration. These findings highlight how maintaining the migratory state of T cells may undermine treatment effectiveness.

The tool represents a significant advance for precision medicine, offering the potential to reduce side effects, optimize costs (such as the average BRL 10,000 per month cost in Brazil’s SUS), and improve patient outcomes by guiding therapy decisions. The research team's plans for the future include validating the approach with larger, more diverse patient groups and exploring its application to other diseases. Researchers also aim to make the morphology marker accessible with simpler, lower-cost equipment, broadening its clinical use.

“The project is extremely interesting and innovative,” said Helder Nakaya, senior researcher and the author of the article. “The great insight was to take the image, transform it into numbers, and use this table in machine learning. I’m sure that it’ll now be possible to replicate this type of approach for other diseases and treatments.”

Related Links:
FAPESP