Ultrafast clustering of single-cell flow cytometry data using FlowGrid.

Area of Science:

• Single-cell analysis
• Computational biology
• Immunology

Background:

• Flow cytometry enables quantitative single-cell profiling of cell surface markers, crucial for identifying cell subpopulations and heterogeneity.
• Traditional manual gating for cell identification is subjective and inefficient for large, multidimensional datasets.
• Existing clustering algorithms struggle with scalability for datasets exceeding ten million cells.

Purpose of the Study:

• To develop a scalable and efficient clustering algorithm for large-scale flow cytometry data.
• To address the limitations of existing methods in handling high-dimensional, high-cell-number datasets.

Main Methods:

• Developed FlowGrid, a novel clustering algorithm combining DBSCAN's density-based approach with grid-based scalability.
• Implemented FlowGrid as an open-source Python package.
• Evaluated FlowGrid's performance against state-of-the-art clustering programs.

Main Results:

• FlowGrid demonstrates linear scalability with cell number and is memory efficient.
• Achieved comparable clustering results to existing methods but with significantly reduced processing time.
• Successfully clustered 23.6 million cells in under 12 seconds, outperforming other algorithms that require over 500 seconds or fail.

Conclusions:

• FlowGrid offers an ultrafast and scalable solution for clustering large single-cell flow cytometry datasets.
• The open-source availability of FlowGrid facilitates its adoption in the research community.
• This algorithm significantly advances the analysis of high-dimensional single-cell data.