Accurate cell type annotation for single-cell chromatin accessibility data via contrastive learning and reference...

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Single-cell chromatin accessibility sequencing (scCAS) advances epigenomic heterogeneity insights.
• Existing automatic annotation methods struggle with reference data and novel cell types.

Purpose of the Study:

• To develop a reference-guided automatic annotation method for scCAS data.
• To effectively identify both known and novel cell types in scCAS datasets.

Main Methods:

• Proposed RAINBOW, a novel method based on a contrastive learning framework.
• Incorporated reference data into the annotation process.
• Utilized contrastive learning to characterize cell type heterogeneity.

Main Results:

• RAINBOW demonstrated superior performance in annotating known and novel cell types compared to state-of-the-art methods.
• Effectiveness of incorporating reference data was verified.
• RAINBOW showed robustness to data sparsity and varying numbers of cell types.
• RAINBOW performed well on newly sequenced data and revealed biological implications.

Conclusions:

• RAINBOW offers a significant advancement in automatic cell type annotation for scCAS data.
• The method effectively handles novel cell types and incorporates reference data.
• RAINBOW is expected to aid scCAS data analysis and biological discovery.