A message passing framework for precise cell state identification with scClassify2.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate cell annotation is essential for analyzing single-cell data.
• Existing methods often focus on distinct cell types, neglecting sequential cell populations.
• There is a need for advanced computational tools to address this gap.

Purpose of the Study:

• To introduce scClassify2, a novel computational method for cell annotation.
• To specifically address the identification of adjacent cell states and sequential cell populations.
• To provide a versatile tool applicable to various single-cell data types.

Main Methods:

• Development of scClassify2, a dual-layer architecture incorporating biological knowledge.
• Application of ordinal regression for sequential cell state identification.
• Validation across different single-cell data platforms, including spatial transcriptomics.

Main Results:

• scClassify2 demonstrates competitive performance against state-of-the-art methods.
• The method effectively identifies sequential cell populations, improving annotation accuracy.
• Generalizability shown across single-cell RNA sequencing and spatial transcriptomics data.

Conclusions:

• scClassify2 offers a significant advancement in cell annotation by focusing on sequential populations.
• The tool is robust and applicable to diverse high-throughput biological data.
• A web server is available to facilitate academic research using scClassify2.