Spatial AI for Predicting Cell Fate Transitions and Disease Course in Multiple Sclerosis
Title: Spatial AI for Predicting Cell Fate Transitions and Disease Course in Multiple Sclerosis
DNr: Berzelius-2026-333
Project Type: LiU Berzelius
Principal Investigator: Gonçalo Castelo-Branco <goncalo.castelo-branco@ki.se>
Affiliation: Karolinska Institutet
Duration: 2026-11-05 – 2027-06-01
Classification: 30105
Homepage: https://ki.se/en/mbb/oligointernode
Keywords:

Abstract

Multiple sclerosis (MS) is a neurological disease characterized by autoimmune attack in the central nervous system, targeting oligodendroglia. Using single-cell transcriptomics/epigenomics in the experimental autoimmune encephalomyelitis (EAE) mouse model of MS and in MS patients, our lab has found that oligodendroglia also actively contribute to MS pathogenesis/inflammation. To determine the spatial distribution of the cell types/states in homeostasis and in disease, we have applied In Situ Sequencing (ISS) at distinct times points and CNS regions of the EAE mouse model of MS and MS tissue samples, characterizing the dynamics of lesion evolution. We have developed and implemented spatial epigenomics, transcriptomics and multiomics, obtaining for instance new insights of developmental oligodendrogenesis. This project will use advanced spatial molecular technologies and artificial intelligence to understand multiple sclerosis in a new way. Instead of studying cells only as isolated units, we will examine where they are located in tissue, which neighbouring cells they interact with, and how their genes and regulatory programmes change during disease. By combining information on gene activity, epigenetic regulation, proteins and tissue architecture, we aim to build a detailed map of how healthy brain cells shift into disease-associated states. Our central goal is to develop AI models that can predict how cell identities change over time and how these changes contribute to different forms of disease. The models will be trained on rich datasets from experimental models of multiple sclerosis and homeostatics states. They will help identify which cell types, molecular signals and tissue environments are most important for lesion formation, disease progression and recovery. In the long term, the project could support the discovery of new biomarkers and therapeutic strategies that are better tailored to the biological processes driving multiple sclerosis in individual patients.