Scalable multi-sample single-cell data analysis by Partition-Assisted Clustering and Multiple Alignments of Networks

Area of Science:

• Immunology
• Computational Biology
• Biotechnology

Background:

• Mass cytometry (CyTOF) enables high-dimensional single-cell analysis with numerous markers.
• Generating and analyzing multiple CyTOF samples presents challenges for subpopulation discovery and cross-sample alignment.
• Existing computational methods struggle with the scalability and complexity of large, multi-sample CyTOF datasets.

Purpose of the Study:

• To develop a computational method for automated cell population identification in mass cytometry data.
• To enable accurate alignment of cell subpopulations across multiple CyTOF samples.
• To facilitate the definition of dataset-level cellular states from large-scale CyTOF experiments.

Main Methods:

• Developed Partition-Assisted Clustering and Multiple Alignments of Networks (PAC-MAN) algorithm.
• PAC-MAN performs fast, automatic identification of cell populations.
• The method aligns subpopulations across samples to define dataset-level cellular states.

Main Results:

• PAC-MAN achieves automated cell population discovery comparable to expert manual analysis.
• The algorithm is computationally efficient, handling large-scale CyTOF datasets.
• Enables robust alignment of cellular states across multiple samples.

Conclusions:

• PAC-MAN overcomes limitations in analyzing multi-sample CyTOF data.
• Provides a scalable solution for subpopulation discovery and cross-sample analysis.
• Accelerates insights from large CyTOF datasets in clinical and cancer research.