Discovery of optimal cell type classification marker genes from single cell RNA sequencing data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) generates vast transcriptional data, enabling cell biology insights.
• Challenges exist in identifying cell types and marker genes from large scRNA-seq datasets.
• Machine learning and AI offer solutions for analyzing scRNA-seq data.

Purpose of the Study:

• To introduce NS-Forest version 4.0, an algorithm for scalable, data-driven marker gene selection.
• To enhance marker gene identification for distinguishing closely related cell types.
• To improve efficiency for analyzing large-scale scRNA-seq atlases.

Main Methods:

• Utilizes a random forest machine learning approach.
• NS-Forest v4.0 features a modularized decision tree for comparing marker gene sets.
• Introduces the On-Target Fraction metric to quantify marker exclusivity.

Main Results:

• NS-Forest v4.0 demonstrates superior performance in simulations and real-world data analysis.
• Identifies marker genes with higher On-Target Fraction values for closely related cell types.
• Outperforms other methods in cell type classification accuracy (higher F-beta scores) on human organ datasets.

Conclusions:

• NS-Forest v4.0 provides an efficient and accurate method for marker gene selection in scRNA-seq.
• The algorithm's modularity allows for flexible marker gene evaluation.
• Identified marker genes have potential applications in spatial transcriptomics and biomedical ontologies.