scParadise: tunable, highly accurate multi-level cell type annotation, unknown cell type identification, and modality...

Area of Science:

• Single-cell biology
• Immunology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) data analysis faces challenges in accurate cell type annotation and multi-omic state interpretation.
• Existing methods struggle with sparse and heterogeneous scRNA-seq datasets, limiting reproducible conclusions.

Purpose of the Study:

• To introduce scParadise, an integrated suite of tools designed to transform scRNA-seq data into actionable scientific knowledge.
• To address limitations in cell type annotation, cross-tissue data imputation, and standardized benchmarking.

Main Methods:

• Development of scAdam for multi-level cell type annotation, including identification of unknown cell types.
• Implementation of scEve for cross-tissue imputation of molecular modalities.
• Utilizing scNoah for standardized benchmarking of scRNA-seq analysis tools.

Main Results:

• Correction of annotation errors in the Tabula Muris Senis atlas, clarifying granulocyte and macrophage identities.
• Identification of three novel natural killer T (NKT) cell subsets in human visceral adipose tissue via protein expression imputation.
• Verification of novel NKT cell subsets using flow cytometry.

Conclusions:

• scParadise significantly improves the accuracy and interpretability of scRNA-seq data.
• The newly identified NKT cell subsets reveal a new paradigm in immune-stromal interactions relevant to chronic inflammation and obesity.
• This work provides a powerful framework for advancing single-cell data analysis and biological discovery.