AI tool helps researchers find new clues about arthritis, dementia and more

As powerful spatial biology technologies allow researchers to measure tens to thousands of genes and proteins across individual cells and map their locations within tissues, researchers need better ways to identify patterns in these complex datasets that are linked to disease.

To address this challenge, a group of researchers, led by Yakir Reshef, PhD, and Soumya Raychaudhuri, PhD, of Mass General Brigham and the Broad Institute of MIT and Harvard, developed a new artificial intelligence-based tool called VIMA.

VIMA helps scientists study tissues in greater detail by analyzing where genes, proteins and cells are located. It uses machine learning to identify small groups of similar tissue regions, known as “microniches,” and determine whether they are associated with disease. Unlike many existing methods, VIMA can pick up on subtle biological signals without requiring researchers to define in advance which cell types or tissue features they want to find.

The team tested the tool using tissue samples from people with rheumatoid arthritis, ulcerative colitis and dementia. Across all three diseases, VIMA identified disease-related tissue features and often detected patterns that were missed by other leading analysis methods.

In an analysis of brain data, VIMA identified a previously unknown type of brain tissue region that was more common in people with dementia than in controls. These regions contained high numbers of oligodendrocytes — cells that help support and insulate nerve fibers — suggesting those cells may play a role in Alzheimer’s disease and related dementias.

The findings show how artificial intelligence can help researchers uncover hidden features of disease in complex spatial molecular datasets. The authors say VIMA could help researchers analyze such datasets across a wide range of conditions, from inflammatory diseases to neurodegenerative disorders.

Published in Nature Methods on October 6, 2026| Read the paper: “Accurate and Well-Powered Case-Control Analysis of Spatial Molecular Data”

Summary reviewed by: Yakir Reshef, PhD, lead author; Soumya Raychaudhuri, PhD, senior author