scSwinTNet: A Cell Type Annotation Method for Large-Scale Single-Cell RNA-Seq Data Based on Shifted Window Attention.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Cell type annotation from single-cell RNA sequencing (scRNA-seq) data is crucial for understanding biological processes.
• Current unsupervised clustering methods require time-consuming and non-repeatable manual annotation.
• The exponential growth of sequencing data necessitates automated and scalable annotation solutions.

Purpose of the Study:

• To develop an automated and accurate tool for cell type annotation in scRNA-seq data.
• To address the limitations of manual annotation, including time, bias, and scalability.
• To integrate large-scale scRNA-seq datasets for improved annotation accuracy.

Main Methods:

• Introduction of scSwinTNet, a pre-trained tool utilizing self-attention with shifted windows.
• Application of intelligent information extraction from gene expression data.
• Validation using a large dataset of 399,760 cells from human and mouse tissues.

Main Results:

• scSwinTNet demonstrates effectiveness and robustness in cell type annotation.
• The tool accurately annotates cell types without requiring prior knowledge or manual intervention.
• Achieved accurate cell type identification across diverse human and mouse tissues.

Conclusions:

• scSwinTNet is the first model to employ a pre-trained shifted window attention mechanism for scRNA-seq cell type annotation.
• The proposed method significantly improves efficiency and accuracy in large-scale single-cell data analysis.
• scSwinTNet offers a scalable and reliable solution for automated cell type identification.