Accurate feature selection improves single-cell RNA-seq cell clustering.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Cell clustering is a key step in single-cell RNA sequencing (scRNA-seq) data analysis.
• Feature selection, identifying informative genes, significantly impacts clustering accuracy.
• Current methods often use simple statistical approaches for feature selection.

Purpose of the Study:

• To evaluate the impact of feature selection on cell clustering accuracy in scRNA-seq data.
• To develop an improved feature selection algorithm for more representative gene subsets.
• To enhance downstream clustering performance using the developed algorithm.

Main Methods:

• Systematic evaluation of feature selection's impact on scRNA-seq cell clustering.
• Development of the FEAture SelecTion (FEAST) algorithm for gene selection.
• Application and validation of FEAST on 12 public scRNA-seq datasets.

Main Results:

• Feature selection quality critically influences cell clustering outcomes.
• The FEAST algorithm identifies more representative genes compared to existing methods.
• Utilizing FEAST-selected features significantly improves clustering accuracy across multiple datasets.

Conclusions:

• Effective feature selection is paramount for reliable cell clustering in scRNA-seq.
• FEAST offers a superior approach to gene selection, enhancing biological insights.
• The FEAST algorithm represents a significant advancement for scRNA-seq data analysis tools.