Flow-SNE: A New Approach for Flow Cytometry Clustering and Visualization

Abstract

Flow cytometry is a technology by which the expression of multiple cellular markers are measured simultaneously for each cell. Analysis of the extracted cytometry dataset is invaluable for biologists in many applications such as identification of various cell types with specific phenotypic properties. Specifically, identification of rare subpopulation in presence of infectious diseases can reveal the effects of that disease on immune system reflex. In this paper, we propose a new automatic clustering technique to extract meaningful cell subtypes from the flow cytometry datasets. In contrast with other methods, our approach is able to visualize the dataset in 2D. Also, it is designed to easily handle very large flow cytometry datasets. We compared our method against others using three famous publicly available flow cytometry datasets. The results indicate a much better performance and an effective visualization.