Nested Stochastic Block Models applied to the analysis of single cell data.

Area of Science:

• Molecular Biology
• Computational Biology
• Bioinformatics

Background:

• Single-cell profiling is crucial for understanding heterogeneous biological systems.
• Clustering cells is a primary endpoint to define functional properties of cell mixtures.
• Current clustering methods often rely on graph-based community detection and modularity optimization.

Purpose of the Study:

• To propose an alternative and principled approach for cell clustering in single-cell analysis.
• To demonstrate the utility of Stochastic Block Models (SBMs) for cell group identification.
• To establish SBMs as a framework for additional single-cell analysis tasks, including label transfer.

Main Methods:

• Application of Stochastic Block Models (SBMs) for analyzing single-cell data.
• Development of a Python library named 'schist' for implementing SBMs.
• Ensuring compatibility of the 'schist' library with the 'scanpy' framework.

Main Results:

• SBMs are shown to be suitable for identifying distinct cell groups.
• The SBM framework effectively supports tasks like label transfer in single-cell datasets.
• The 'schist' library provides a user-friendly tool for applying SBMs.

Conclusions:

• Stochastic Block Models offer a principled and effective alternative for cell clustering.
• SBMs provide a versatile framework for various single-cell analysis tasks beyond clustering.
• The 'schist' Python library facilitates the adoption of SBMs in the research community.