Recursive Clustering of Cellular Diversity in scRNA-Seq Data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Standard single-cell RNA sequencing (scRNA-seq) clustering uses a single feature extraction and clustering round.
• This can miss rare cell types and subtle phenotypic differences due to non-specific gene features.
• Identifying distinct cell populations is crucial for understanding biological systems and disease states.

Purpose of the Study:

• To develop and evaluate a recursive feature extraction and clustering approach for scRNA-seq data.
• To improve the detection of cell types, especially rare ones, and capture finer phenotypic distinctions.
• To provide a more granular and interpretable hierarchical framework for cell clustering.

Main Methods:

• A recursive clustering algorithm was designed, involving iterative feature extraction and clustering on existing clusters.
• The recursive method was benchmarked against conventional non-recursive clustering using four scRNA-seq datasets.
• The approach was applied to scRNA-seq data from Crohn's disease patients with three clinical phenotypes.

Main Results:

• The recursive clustering approach demonstrated robust improvement in cell type detection across various resolution parameters.
• It successfully identified phenotypic differences in Crohn's disease data not apparent with conventional methods.
• The method revealed cell clusters within gene expression feature spaces specific to each cluster, enhancing interpretability.

Conclusions:

• Recursive clustering offers a significant advancement over traditional methods for scRNA-seq data analysis.
• This approach enhances the resolution and accuracy of cell type identification, particularly for complex and rare cell populations.
• The hierarchical framework provided by recursive clustering aids in understanding disease heterogeneity and biological variation.