CTISL: a dynamic stacking multi-class classification approach for identifying cell types from single-cell RNA-seq...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate cell type identification is crucial for single-cell RNA sequencing (scRNA-seq) data analysis.
• Existing supervised machine learning predictors often use single classifiers, limiting performance.
• There is a need for more accurate computational models for robust cell type identification.

Purpose of the Study:

• To develop an ensemble learning strategy for improved cell type identification in scRNA-seq data.
• To introduce CTISL, a two-layer stacking model integrating multiple classifiers.
• To enhance the accuracy and robustness of cell type classification.

Main Methods:

• CTISL employs a two-layer stacking ensemble learning approach.
• The first layer combines multiple cell-type-specific classifiers (e.g., support-vector machine, logistic regression) as base learners.
• A meta-classifier in the second layer integrates the base learner outputs for final cell type prediction.

Main Results:

• CTISL was evaluated across 24 benchmarking experiments on 17 human and mouse scRNA-seq datasets.
• The model demonstrated superior or competitive performance compared to existing state-of-the-art predictors.
• CTISL provides accurate and reliable cell type identification.

Conclusions:

• CTISL offers a powerful ensemble learning approach for scRNA-seq data analysis.
• The developed model enhances the accuracy and robustness of cell type identification.
• CTISL is a valuable tool for cost-effective cell type identification from scRNA-seq datasets.