scPANDA: PAN-Blood Data Annotator with a 10-Million Single-Cell Atlas

Area of Science:

• Hematology
• Immunology
• Bioinformatics
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) advances cellular heterogeneity studies, especially in the hematological system.
• Accurate immune cell type annotation is crucial but challenging due to cell complexity.
• Existing methods require improvement for precise blood cell identification.

Purpose of the Study:

• To develop the PAN-blood single-cell Data Annotator (scPANDA) for precise blood cell type annotation.
• To leverage a comprehensive 10-million-cell atlas as a reference for scRNA-seq data.
• To overcome challenges in annotating complex immune cell populations.

Main Methods:

• Constructed a 10-million-cell atlas from 16 studies with rigorous quality control and integration.
• Employed scPANDA with a three-layer inference approach for progressive cell type refinement.
• Utilized iterative clustering and harmonization for cell type purity and evaluated performance on external datasets.

Main Results:

• The atlas features a hierarchical structure: 16 compartments, 54 classes, 4,460 low-level clusters, and 611 high-level clusters.
• scPANDA demonstrated robust performance in annotating diverse immune scRNA-seq datasets.
• The tool successfully analyzed immune-tumor coexisting clusters and identified conserved cell clusters across species.

Conclusions:

• scPANDA represents an effective reference mapping strategy using a large-scale atlas.
• The tool significantly enhances the accuracy and reliability of blood cell type identification.
• This approach advances the study of cellular heterogeneity in hematological and immunological research.