ScInfeR: an efficient method for annotating cell types and sub-types in single-cell RNA-seq, ATAC-seq, and spatial...

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Cell-type annotation is crucial for single-cell and spatial omics but faces challenges with limited high-quality references and poor performance in existing methods, especially for scATAC-seq and spatial transcriptomics.
• Current annotation approaches often rely solely on single-cell RNA sequencing (scRNA-seq) references or predefined marker sets, leading to potential bias and usability issues due to data scarcity.

Purpose of the Study:

• To develop a versatile and accurate graph-based cell-type annotation method, ScInfeR, that overcomes limitations of existing tools by integrating multiple data sources.
• To enhance cell-type and subtype identification across diverse omics datasets, including scRNA-seq, scATAC-seq, and spatial transcriptomics, while addressing data scarcity and batch effects.

Main Methods:

• ScInfeR employs a graph-based approach with a hierarchical framework, inspired by graph neural networks, to combine scRNA-seq references and marker sets for annotation.
• The method is designed to be versatile, incorporating chromatin accessibility data for scATAC-seq and spatial coordinates for spatial transcriptomics, and supports weighted markers for nuanced classification.
• ScInfeRDB, an accompanying database, provides curated scRNA-seq references and marker sets for 329 cell-types across human and plant tissues.

Main Results:

• ScInfeR demonstrated superior performance in extensive benchmarking across multiple atlas-scale datasets, outperforming 10 existing tools in over 100 cell-type prediction tasks.
• The method showed robustness against batch effects commonly found in omics datasets.
• ScInfeR accurately annotates a broad range of cell-types and subtypes across scRNA-seq, scATAC-seq, and spatial omics data.

Conclusions:

• ScInfeR offers a robust and accurate solution for cell-type annotation in single-cell and spatial omics, integrating diverse data types effectively.
• The tool's versatility and performance improvements address key challenges in the field, facilitating broader application of omics data analysis.
• The public availability of ScInfeR and its database (ScInfeRDB) promotes reproducible research and advances cell-type annotation capabilities in genomics.