Cell Type Annotation Model Selection: General-Purpose vs. Pattern-Aware Feature Gene Selection in Single-Cell RNA-Seq...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput sequencing enables cell heterogeneity research.
• Gene expression profiles determine cell functionality.
• Manual cell cluster annotation is a bottleneck.

Purpose of the Study:

• To compare XGBoost and Support Vector Machine (SVM) for scRNA-seq data analysis.
• To evaluate the effectiveness of information gain for feature selection.
• To assess scalability and automation in cell type identification.

Main Methods:

• Comparative analysis of XGBoost and SVM.
• Utilized information gain for feature selection.
• Experiments conducted on three standard scRNA-seq datasets.

Main Results:

• XGBoost provides simpler and more scalable automatic cell type annotation.
• XGBoost outperformed SVM in tested scenarios.
• Feature selection enhanced classifier performance.

Conclusions:

• XGBoost is a promising method for automated cell type identification in scRNA-seq data.
• Boosting tree approaches combined with deep neural networks show potential for scRNA-seq analysis.
• This approach can aid in marker gene identification and other biological studies.