CeiTEA: Adaptive Hierarchy of Single Cells with Topological Entropy

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity.
• Existing clustering methods struggle with complex hierarchical cell type structures.

Purpose of the Study:

• Introduce CeiTEA, a novel adaptive hierarchical clustering algorithm.
• Address limitations of current methods in scRNA-seq data analysis.
• Improve the representation of cell type relationships and diversity.

Main Methods:

• CeiTEA employs adaptive hierarchical clustering based on topological entropy (TE).
• It constructs a multi-nary partition tree by minimizing TE.
• Combines bottom-up hierarchy construction with top-down local diversification.

Main Results:

• CeiTEA demonstrated superior clustering performance on simulated and real scRNA-seq data.
• Achieved lower average clustering information distance (0.15) on simulated data compared to SEAT (0.39) and traditional methods (0.67).
• Identified biologically relevant hierarchies reflecting developmental potency in real datasets.

Conclusions:

• CeiTEA effectively captures complex hierarchical structures in single-cell data.
• It offers a powerful tool for analyzing cellular heterogeneity and relationships.
• Potential applications include tumor heterogeneity and tissue specification studies.