Unsupervised Clustering of Cell Populations in Germinal Centers Using Multiplexed Immunofluorescence.

Abstract

Follicles (Fs)/Germinal Centers (GCs) in tonsils and lymph nodes are dynamic microenvironments where diverse immune cell populations interact for the development of antibody responses against pathogens. The accurate in situ phenotypic analysis of these immune cells is a prerequisite for the comphehensive understanding of GC development. In this study, we explore unsupervised clustering approaches for distinguishing cell populations within F/GCs using marker expression data. We evaluate multiple clustering algorithms and find that k-means clustering provides the most effective separation of distinct cell subsets. Additionally, we investigate the predictive potential of common GC markers (CD3, CD4, CD20 and BCL6) for PD-1 expression, an important immune checkpoint regulator. Our analysis demonstrates that PD-1 expression can be reliably inferred using these markers, suggesting potential applications for automated cell classification in immunological studies. This approach enhances our ability to analyze immune cell heterogeneity and may contribute to improved understanding of GC dynamics in health and disease. Our findings support the use of computational clustering for high-dimensional immune profiling.