A Clustering Method for Single-Cell RNA-Seq Data Based on Automatic Weighting Penalty and Low-Rank Representation

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput single-cell RNA sequencing (scRNA-seq) provides detailed cell expression data.
• Clustering analysis is vital for identifying cell identity in scRNA-seq data.
• Existing clustering methods face challenges due to data characteristics and ignore feature contributions, leading to information loss.

Purpose of the Study:

• To develop an improved clustering method for scRNA-seq data.
• To address limitations of existing methods that treat all features equally.
• To enhance the accuracy and stability of cell clustering.

Main Methods:

• Introduction of a weighted distance constraint in similarity graph construction.
• Integration of a similarity constraint for a more symmetric affinity matrix.
• Proposal of the Joint Automatic Weighting Similarity Graph and Low-rank Representation (JAGLRR) method.

Main Results:

• JAGLRR effectively evaluates feature contributions, prioritizing significant features and reducing redundant ones.
• The method recovers data's linear relationships more accurately and extracts more discriminative information.
• JAGLRR demonstrated superior performance over 11 existing methods on simulated and real scRNA-seq datasets.

Conclusions:

• JAGLRR offers a significant advancement in scRNA-seq data clustering.
• The method achieves higher accuracy and stability compared to current approaches.
• JAGLRR provides more reliable cell identity identification for biological research.