SSRE: Cell Type Detection Based on Sparse Subspace Representation and Similarity Enhancement.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate cell type identification is crucial for single-cell RNA sequencing (scRNA-seq) data analysis.
• Current unsupervised clustering methods for scRNA-seq data often struggle with accuracy due to limitations in similarity measurement.
• Improving cell type identification is essential for advancing various scRNA-seq studies.

Purpose of the Study:

• To develop a novel single-cell clustering framework, SSRE, that enhances the accuracy of cell type identification from scRNA-seq data.
• To leverage similarity learning and sparse representation for robust cell-to-cell relationship modeling.
• To provide a versatile tool applicable to scRNA-seq data visualization and differential gene expression analysis.

Main Methods:

• Proposed SSRE (Single-cell Similarity Representation Ensemble) framework utilizing similarity learning.
• Modeled cell relationships based on subspace assumptions to generate sparse cell-to-cell similarity representations.
• Integrated three classical pairwise similarities with a gene selection and enhancement strategy.

Main Results:

• SSRE demonstrated superior performance in cell type identification across ten real and five simulated scRNA-seq datasets compared to state-of-the-art methods.
• The sparse representation effectively retained the most similar neighbors for each cell, improving clustering outcomes.
• The framework showed potential for extending scRNA-seq data visualization and differentially expressed gene identification.

Conclusions:

• SSRE offers a significant advancement in unsupervised clustering for scRNA-seq data, leading to more accurate cell type identification.
• The method's effectiveness is validated across diverse datasets, highlighting its robustness.
• SSRE provides a valuable tool for the broader scRNA-seq research community, with implementations available in MATLAB and Python.