scASK: A Novel Ensemble Framework for Classifying Cell Types Based on Single-cell RNA-seq Data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• The Human Cell Atlas (HCA) project aims to map all human cell types.
• Single-cell RNA sequencing (scRNA-seq) is crucial for HCA, relying on dimension reduction and clustering.
• Current computational methods struggle with large-scale scRNA-seq data, exhibiting poor performance and sensitivity to initial data.

Purpose of the Study:

• To introduce a novel ensemble framework, Adaptive Slice KNNs (scASK), for analyzing high-dimensional scRNA-seq data.
• To address the limitations of existing computational approaches in terms of performance and data sensitivity.

Main Methods:

• Developed scASK, an ensemble framework comprising three modules: Data Adaptive Slicing (DAS), Meta Classifiers Selecting (MCS), and Ensemble Mode Switching (EMS).
• Evaluated scASK using thirteen real-world scRNA-seq datasets.
• Compared scASK performance against five popular classification algorithms.

Main Results:

• scASK demonstrated superior accuracy and robustness compared to five established classification algorithms across thirteen scRNA-seq datasets.
• The Data Adaptive Slicing (DAS) module proved effective for structural reduction.
• The ensemble framework (scASK) provides a novel and robust approach for cell type classification.

Conclusions:

• Adaptive slicing is an effective procedure for structural reduction in scRNA-seq data analysis.
• scASK offers a robust and novel ensemble framework for accurate cell type classification from scRNA-seq data.
• scASK is publicly available for use in cell type identification studies.