scWizard: A web-based automated tool for classifying and annotating single cells and downstream analysis of...

Area of Science:

• Computational Biology
• Cancer Research
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-Seq) generates vast datasets for cancer cell type characterization.
• Existing tools lack effective integration for fine-grained annotation of tumor microenvironment (TME) cell subtypes.
• Accurate TME cell annotation is crucial for advancing cancer research.

Purpose of the Study:

• To develop scWizard, a user-friendly, point-and-click tool for automated scRNA-Seq analysis in cancer research.
• To enhance the accuracy and efficiency of cell type and subtype annotation within the TME.
• To provide a comprehensive platform for downstream analyses and visualization of cancer scRNA-Seq data.

Main Methods:

• Developed scWizard, a tool integrating automated cell annotation using deep neural network learning and 11 downstream analyses.
• Utilized a reference dataset of 113,976 cells across 13 cancer types for hierarchical model training.
• Incorporated a pre-trained dataset for flexible user choice and comparative analysis against existing methods.

Main Results:

• scWizard automatedly classifies and annotates 7 major cell types and 47 cell subtypes in the TME.
• Achieved higher accuracy in annotating tumor-derived T/NK and myeloid cell subtypes compared to five existing methods.
• Demonstrated robustness across three independent cancer datasets with high accuracy for major cell types (0.98) and subtypes (0.85 for myeloid, 0.79 for T/NK).

Conclusions:

• scWizard offers a robust, accurate, and user-friendly solution for cancer scRNA-Seq data analysis, focusing on TME decoding.
• The tool's automated analysis and visualization capabilities accelerate biological discovery in cancer research.
• scWizard is widely applicable across different cancer types and cell populations, facilitating broader research use.