Random forest based similarity learning for single cell RNA sequencing data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is a powerful tool for biological and biomedical research.
• Accurate cell-cell similarity assessment is crucial for cell type discovery and characterization in scRNA-seq studies.
• Existing analysis methods may be suboptimal for scRNA-seq data due to noise and specific data characteristics.

Purpose of the Study:

• To develop and present RAFSIL, a random forest-based approach for learning cell-cell similarities from scRNA-seq data.
• To provide a flexible and expandable tool for scRNA-seq data analysis.
• To improve the accuracy of cell type identification and the detection of technical variation.

Main Methods:

• RAFSIL employs a two-step procedure: scRNA-seq-specific feature construction followed by similarity learning.
• The method utilizes a random forest algorithm for robust similarity estimation.
• RAFSIL is designed for adaptability and can be integrated into various downstream analyses.

Main Results:

• RAFSIL demonstrates superior performance compared to existing methods across diverse scRNA-seq datasets.
• The approach effectively identifies and highlights unwanted technical variations in scRNA-seq data.
• RAFSIL similarities facilitate improved dimension reduction, visualization, and clustering of single cells.

Conclusions:

• RAFSIL offers a flexible and effective solution for learning cell-cell similarities from scRNA-seq data.
• The tool enhances the accuracy of cell type discovery and the identification of data artifacts.
• RAFSIL represents a valuable advancement in scRNA-seq data analysis, improving biological insights.