RFCell: A Gene Selection Approach for scRNA-seq Clustering Based on Permutation and Random Forest.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity in biology and medicine.
• Clustering is a key analysis method for scRNA-seq data, aiding in cell population discovery and developmental trajectory inference.
• Effective gene selection is vital for improving scRNA-seq clustering accuracy and reducing data dimensionality.

Purpose of the Study:

• To enhance scRNA-seq data clustering through improved gene selection.
• To introduce RFCell, a novel supervised gene selection method.
• To evaluate RFCell's performance against existing gene selection techniques.

Main Methods:

• Proposed RFCell, a supervised gene selection method utilizing permutation and random forest classification.
• Applied RFCell and three other gene selection methods to 10 scRNA-seq datasets.
• Utilized three classical clustering algorithms to assess the performance of selected gene sets.

Main Results:

• RFCell demonstrated superior gene selection performance compared to three existing methods.
• Gene selection using RFCell contributed to improved cell type identification.
• The method effectively reduced dimensionality while enhancing clustering accuracy.

Conclusions:

• RFCell is an effective method for supervised gene selection in scRNA-seq data analysis.
• Improved gene selection significantly enhances the accuracy of cell type identification and clustering.
• RFCell offers a valuable tool for advancing scRNA-seq data analysis in biological and medical research.