An automatic annotation tool and reference database for T cell subtypes and states at single-cell resolution

Area of Science:

• Immunology
• Genomics
• Computational Biology

Background:

• T cells exhibit diverse subtypes and functional states crucial for immune responses.
• Lack of comprehensive reference atlases and automated annotation tools hinders T cell research.
• Analyzing T cell heterogeneity is vital for understanding various physiological and pathological conditions.

Purpose of the Study:

• To construct the largest human T cell reference dataset to date.
• To develop an automated tool (STCAT) for accurate T cell subtype and state annotation from scRNA-seq data.
• To facilitate deeper insights into T cell immunity in cancer and infectious diseases.

Main Methods:

• Compiled a reference dataset of 1,348,268 T cells from 35 conditions and 16 tissues.
• Classified T cells into 33 subtypes and 68 categories using hierarchical models and marker correction.
• Developed and validated the STCAT tool on independent scRNA-seq datasets, comparing its accuracy to existing methods.

Main Results:

• STCAT demonstrated 28% higher accuracy than existing tools in T cell annotation.
• Identified enrichment of CD4+ Th17 cells in late-stage lung cancer and MAIT cells in mild COVID-19.
• Revealed decreased Treg cytotoxicity in post-treatment ovarian cancer and distinct CD4+/CD8+ T cell distributions in tumor vs. healthy samples.

Conclusions:

• The developed human T cell reference atlas and STCAT tool significantly enhance T cell research capabilities.
• These resources provide a foundation for analyzing T cell immunity in diverse conditions, including cancer immunology.
• The TCellAtlas database offers a platform for exploring T cell expression and analyzing custom scRNA-seq data.