scDetect

scDetect classifies cell types in single-cell RNA sequencing (scRNA-seq) datasets to improve identification of tumor and non-tumor cells across batches and sequencing platforms.

Key Features:

• Rank-Based Ensemble Learning Algorithm: Employs a rank-based ensemble approach combining gene expression rank analysis with a probability-based majority-vote ensemble for cell type prediction.
• Cross-Platform Compatibility: Addresses batch effects and sequencing platform variability to provide consistent classification across different scRNA-seq studies.
• Incorporation of Tumor-Specific Features: Integrates tumor-specific features including cell copy number variation consensus clustering and epithelial score into the classification framework.
• High Accuracy in Classification: Demonstrated superior accuracy on datasets from pancreatic tissue, mononuclear cells, and tumor biopsies compared with other publicly available tools.

Scientific Applications:

• Single-cell RNA profiling: Enhances precise cell type identification in scRNA-seq experiments.
• Cancer research: Enables detection and characterization of tumor cells within heterogeneous tumor biopsies using copy number and epithelial features.
• Cross-study integration: Facilitates comparative analyses across studies and platforms by mitigating batch and platform variability.

Methodology:

Performs gene expression rank-based analysis, a majority-vote ensemble machine-learning probability-based prediction, and integrates tumor-specific features such as cell copy number variation consensus clustering and epithelial score.