High-precision cell-type mapping and annotation of single-cell spatial transcriptomics with STAMapper.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• Single-cell spatial transcriptomics (scST) maps gene expression within tissue context.
• Integrating scRNA-seq and scST data is crucial for understanding tissue architecture and cell function.

Purpose of the Study:

• To develop a computational method for transferring cell-type labels from scRNA-seq to scST data.
• To enhance cell-type annotation accuracy in scST datasets.
• To enable the discovery of novel cell types and subtypes within spatial transcriptomics data.

Main Methods:

• Development of STAMapper, a heterogeneous graph neural network.
• Training and validation using 81 scST datasets and 16 paired scRNA-seq datasets.
• Benchmarking against existing computational methods for cell-type transfer.

Main Results:

• STAMapper achieved superior performance on 75 out of 81 scST datasets.
• The method demonstrated enhanced accuracy, especially at cell cluster boundaries.
• STAMapper successfully identified previously unknown cell types and provided precise subtype annotations.

Conclusions:

• STAMapper is an effective tool for cell-type annotation in scST data.
• The method improves upon manual annotation and enables deeper biological insights.
• This approach facilitates the integration of diverse single-cell genomics datasets.