gCAnno: a graph-based single cell type annotation method.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Current single-cell RNA analysis often annotates cell types at the cluster level, not the individual cell level.
• Clustering methods and parameters significantly impact annotation accuracy, necessitating iterative refinement.
• Reference atlas-based methods show promise but lack robustness across different data noise levels and platforms.

Purpose of the Study:

• To develop a robust and accurate method for single-cell type annotation.
• To overcome limitations of existing cluster-level annotation and reference-based approaches.

Main Methods:

• gCAnno constructs a cell type-gene bipartite graph and uses graph embedding to identify cell type-specific genes.
• Naïve Bayes (gCAnno-Bayes) and Support Vector Machine (gCAnno-SVM) classifiers are employed for annotation.
• The method was evaluated on multiple single-cell datasets with varying noise and platform origins.

Main Results:

• gCAnno demonstrated superior performance compared to state-of-the-art methods.
• The tool achieved higher accuracy and robustness in cell type annotation.
• Performance was consistent across datasets with different noise levels and from diverse platforms.

Conclusions:

• gCAnno offers a robust and accurate solution for single-cell RNA analysis cell type annotation.
• The source code is publicly available, facilitating wider adoption and research.