P09 (2026 - 2030) Application of artificial intelligence for the analysis of multi-parameter cytometric data sets of rare B cell subsets

Hauke Busch (UzL), Rudolf Manz (UzL), Mareike Ohms (UzL)

Project Summary 

The recent advances in multi-parameter, flow cytometry, scRNAseq, spatial transcriptomics and other high throughput methods currently revolutionize biomedical research. State-of-the-art flow cytometers can analyze up to 40 parameters simultaneously for each individual cell and allow acquisition of a million of cells per sample, a pre-requisite for the detection and characterization of populations of rare cell types, such as B cell and plasma cell subpopulations. Nevertheless, this method has practical limitations due to missing methods for accurate automatic compensation and identification of these rare cell types. Consequently, establishing flow cytometry settings and analysis gates for larger flow cytometric panels of 20 or more parameters requires extensive manual intervention and expert knowledge. Hence, a major limitation of multi-parameter flow cytometry and other high throughput methods is the complexity of the derived data sets, which currently require time consuming and labor-intensive analysis. Deep learning methods can provide novel and promising approaches for the analysis of big data sets derived in a variety of high throughput approaches. 

This project will develop machine-based learning to assist or even automate panel generation and multi-parameter analysis of rare B cell/plasma cell subsets by flow cytometry, with the perspective to adopt similar approaches later for analyzing (spatial) single cell transcriptomics/proteomics data. The project will closely collaborate with other projects of RTG 3095 that will provide the samples and the practical experience with manual data analysis, required to train the algorithms.