ScanNet: Single-cell annotation informed by transcriptional regulation Network via iterative heterogeneous graph...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Accurate cell type annotation is crucial for understanding cellular functions in single-cell RNA sequencing (scRNA-seq) data.
• Existing methods often fail to fully utilize the rich information within transcriptional regulatory networks (TRNs).
• TRNs map regulatory relationships between transcription factors (TFs) and target genes (TGs), providing insights into transcriptional programs.

Purpose of the Study:

• To introduce ScanNet, a novel method for single-cell annotation that leverages prior knowledge of TRNs.
• To integrate gene expression data with TRN mechanisms for improved cell type identification.
• To capture cell-type-specific regulatory characteristics for enhanced annotation accuracy.

Main Methods:

• ScanNet employs an iterative heterogeneous graph convolutional framework with a dual-channel encoder.
• The Regulation-level Encoder captures local TF-TG interactions within the TRN graph structure.
• The Expression-level Encoder learns global cellular transcriptional states from gene expression data.

Main Results:

• ScanNet consistently outperforms ten state-of-the-art cell type annotation methods across eight diverse scRNA-seq datasets.
• The method demonstrates robust performance in cross-platform annotation and identifying novel cell types.
• ScanNet achieves superior results when adapted for single-cell ATAC-seq (scATAC-seq) data.

Conclusions:

• ScanNet is a scalable, transferable, and mechanistically informed framework for accurate cell type annotation.
• The integration of TRN information enhances the understanding of cell-type-specific regulatory mechanisms.
• The framework's adaptability to different single-cell data modalities, including scATAC-seq, highlights its broad applicability.