Accurate Annotation for Differentiating and Imbalanced Cell Types in Single-Cell Chromatin Accessibility Data.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Single-cell chromatin accessibility sequencing (scCAS) reveals epigenomic heterogeneity.
• Automatic cell type annotation is crucial but challenging for scCAS data.
• Existing methods struggle with differentiating and imbalanced cell types.

Purpose of the Study:

• To develop a robust computational method for accurate cell type annotation in scCAS data.
• To address limitations of current methods in handling noisy and imbalanced cell populations.
• To improve the characterization of cell heterogeneity using scCAS data.

Main Methods:

• Proposed CASCADE, a novel annotation method.
• Employed simulation- and denoising-based strategies.
• Conducted comprehensive experiments on multiple scCAS datasets.

Main Results:

• CASCADE effectively distinguishes cell types and mitigates noise.
• Achieved superior annotation performance for differentiating and imbalanced cell types.
• Demonstrated robustness to batch effects, data sparsity, and varying numbers of cell types.
• Outperformed baseline methods on newly sequenced data.

Conclusions:

• CASCADE offers a significant advancement in scCAS data analysis.
• The method enhances the accurate annotation of cell types, especially in complex datasets.
• CASCADE is expected to facilitate deeper understanding of cell heterogeneity.