Deciphering cell types by integrating scATAC-seq data with genome sequences.

Area of Science:

• Genomics and Epigenetics
• Computational Biology
• Single-Cell Analysis

Background:

• Single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) offers insights into gene regulation and epigenetic heterogeneity.
• Accurate cell annotation from scATAC-seq data is challenging due to high dimensionality and data sparsity.
• Current methods often neglect the underlying genomic sequence information.

Purpose of the Study:

• To develop an accurate single-cell annotation method for scATAC-seq data.
• To integrate genomic sequence information around accessibility peaks into the annotation process.
• To improve cell identification and uncover functional genomic elements.

Main Methods:

• Proposed SANGO method integrates genome sequences around accessibility peaks in scATAC data.
• Genomic sequences are encoded into low-dimensional embeddings and used to reconstruct cell peak statistics.
• A graph transformer network aligns query and reference cells for annotation, utilizing learned regulatory modes.

Main Results:

• SANGO consistently outperformed competing methods across 55 paired scATAC-seq datasets.
• The method successfully identified unknown tumor cells using attention edge weights.
• Cell-type-specific peaks were identified, providing functional insights via enrichment analyses.

Conclusions:

• SANGO provides a robust and accurate approach for single-cell annotation using scATAC-seq data.
• Integrating genomic sequence information enhances cell annotation accuracy and biological discovery.
• The method has potential for identifying novel cell types and understanding regulatory mechanisms.