flowPeaks: a fast unsupervised clustering for flow cytometry data via K-means and density peak finding

Area of Science:

• Computational Biology
• Bioinformatics
• Data Science

Background:

• Unsupervised clustering of flow cytometry data faces challenges with existing methods.
• Finite mixture models are slow and struggle with irregular cluster shapes.
• Histogram spatial exploration is unsuitable for high-dimensional data due to computational and memory constraints.

Purpose of the Study:

• To develop a novel algorithm for unsupervised clustering of high-dimensional flow cytometry data.
• To overcome the limitations of existing clustering approaches in terms of speed, scalability, and cluster shape identification.

Main Methods:

• The flowPeaks algorithm combines finite mixture models and histogram spatial exploration.
• It uses K-means clustering for initial data partitioning.
• A smoothed density function is generated, and local peaks are identified to cluster cells.

Main Results:

• flowPeaks is applicable to high-dimensional data and identifies irregular cluster shapes.
• The algorithm is automatic, fast, reliable, and robust to outliers.
• Performance was validated against established algorithms like Misty Mountain, FLOCK, flowMeans, flowMerge, and FLAME.

Conclusions:

• flowPeaks offers an effective solution for unsupervised clustering in flow cytometry.
• The algorithm provides a robust and efficient method for analyzing complex, high-dimensional datasets.
• The R package flowPeaks is publicly available for use.