Identifying Pathogenic B Cells in Multiple Sclerosis
Machine Learning-Driven Meta-Analysis Reveals Shared B Cell Repertoire linked to Multiple Sclerosis
Abstract
Multiple sclerosis (MS) is an autoimmune, neurodegenerative disease where the immune system damages myelin sheaths in the central nervous system. MS pathogenesis was attributed to auto-reactive T cells. However, recent studies suggest that B cells also play a vital role, including antibody-dependent and independent effects in compartmentalized inflammation. Clonotypes are groups of B cells with the same receptor.
To explore the role of B cell clonotypes in MS pathogenesis, we conducted a meta-analysis of raw high-throughput sequencing data of the B cell repertoire of 82 subjects (48 MS patients and 34 controls). A novel combination of density-based clustering (DBSCAN) and supervised ML (random forest) was applied to cluster clonotypes into unique clonotype groups (UCG), and their disease relevance was assessed. Random forest feature importances, Mann-Whitney tests, and empirical power estimation revealed six significant, potentially pathogenic UCGs.
Additionally, tertiary analyses revealed genetic variability and clonal expansion as hallmarks of MS-associated B cells. Interestingly, new genomic motifs among our potentially pathogenic UCGs aligned with known IGHV4 family biases. Overall, the study highlights the utility of machine learning analysis in decoding immune repertoires and contributes to our understanding of B-cell involvement in MS pathogenesis, paving the way for future therapeutic exploration.
CREST Gold
IRIS National Fair, Top 100
S.T. Yau, Round 2















