Integration of unpaired single cell omics data by deep transfer graph convolutional network

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Large-scale single-cell RNA sequencing (scRNA-seq) and single-cell chromatin accessibility (scATAC-seq) data offer deep biological insights.
• Transferring labels between these modalities is crucial for exploring single-cell omics data.
• Current label transfer methods struggle with fine-grained cell populations and dataset heterogeneity.

Purpose of the Study:

• To present a robust deep transfer model, scTGCN, for accurate label transfer between scRNA-seq and scATAC-seq data.
• To improve the preservation of biological variation during cross-modal label transfer.
• To enable efficient integration of large-scale single-cell omics datasets.

Main Methods:

• Developed a deep transfer model based on graph convolutional networks (GCNs).
• Implemented scTGCN for label transfer between scRNA-seq and scATAC-seq datasets.
• Evaluated performance on mouse atlas data and multimodal APSA-seq and CITE-seq data.

Main Results:

• scTGCN demonstrates versatile performance in preserving biological variation.
• Achieved integration of hundreds of thousands of cells within minutes with low memory consumption.
• Showcased high label transfer accuracy and effective knowledge transfer across different modalities.

Conclusions:

• scTGCN offers a powerful solution for integrating and analyzing multimodal single-cell omics data.
• The model effectively addresses limitations of current label transfer methods.
• Enables robust and efficient exploration of complex biological mechanisms using single-cell omics datasets.