Computational analysis of high-dimensional flow cytometric data for diagnosis and discovery

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Flow cytometry can now measure tens of parameters on millions of cells, creating complex datasets.
• Traditional manual analysis and existing bioinformatics tools struggle to fully analyze these high-dimensional flow cytometry data.

Purpose of the Study:

• To present a general data analysis pipeline for flow cytometry data.
• To enable automatic identification of known cell populations for diagnostic purposes.
• To facilitate exploratory analysis for discovering new biological insights and disease correlates.

Main Methods:

• Overview of a general data analysis pipeline for high-dimensional flow cytometry data.
• Application of unsupervised discovery methods for exploratory analysis.
• Discussion of algorithms for cell population identification (clustering, specific identification) and supervised analysis.

Main Results:

• Demonstration of unsupervised discovery in three real-world basic and clinical research examples.
• Highlights the utility of the proposed pipeline in handling complex flow cytometry datasets.
• Identifies challenges in evaluating algorithms for cell population identification and supervised analysis.

Conclusions:

• The presented data analysis pipeline offers a robust approach for analyzing complex flow cytometry data.
• Unsupervised discovery methods are valuable for both diagnostic and exploratory research in flow cytometry.
• Addressing algorithmic evaluation challenges is crucial for advancing flow cytometry data analysis.