Boosting scRNA-seq data clustering by cluster-aware feature weighting.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) facilitates the study of cell heterogeneity through data clustering.
• Effective clustering relies on selecting relevant genes for evaluating cell similarity.
• Feature selection significantly impacts the effectiveness and efficiency of scRNA-seq data clustering.

Purpose of the Study:

• To introduce CaFew, a novel gene selection method for scRNA-seq data clustering.
• To enhance the performance of existing scRNA-seq clustering approaches through improved feature selection.

Main Methods:

• CaFew employs cluster-aware feature weighting by optimizing the clustering objective function.
• A feature weight matrix is generated to identify important genes for selection.
• Genes with high weights in specific clusters or significant weight variation across clusters are selected.

Main Results:

• CaFew significantly improves clustering performance across 8 real scRNA-seq datasets.
• Combining CaFew with SC3 achieves state-of-the-art results in scRNA-seq clustering.
• CaFew also demonstrates benefits for the visualization of scRNA-seq data.

Conclusions:

• CaFew is an effective method for scRNA-seq data clustering.
• Its gene selection mechanism, based on cluster-aware feature weighting, is key to its success.
• CaFew serves as a valuable tool for scRNA-seq data analysis.