scWECTA

scWECTA implements a soft weighted ensemble classification framework to annotate cell types from single-cell RNA sequencing (scRNA-seq) data.

Key Features:

• Weighted Ensemble Framework: Employs a soft weighted ensemble integrating five distinct classifiers and five informative gene sets to improve annotation robustness.
• Constrained Non-Negative Least Squares: Determines ensemble weights using constrained non-negative least squares (NNLS) optimization.
• Cross-Platform and Tissue Versatility: Validated on multiple pairs of scRNA-seq datasets across platforms and tissues and robust to imbalanced datasets containing rare cell types.
• Balanced Prediction Accuracy: Balances prediction accuracy for common cell types while minimizing the unassigned rate for non-common cell types.

Scientific Applications:

• Cell type annotation: Annotates cell types in single-cell RNA sequencing (scRNA-seq) datasets across diverse platforms and tissues.
• Rare cell population detection: Identifies rare cell populations in imbalanced single-cell datasets.
• Cross-platform comparative analysis: Enables comparative annotation across different scRNA-seq platforms and tissue types.
• Studies of cellular heterogeneity: Supports investigations of tissue heterogeneity in developmental biology, immunology, and oncology.

Methodology:

Integrates five classifiers via a soft weighted ensemble using five informative gene sets, with ensemble weights estimated by constrained non-negative least squares (NNLS).