scGAMNN: Graph Antoencoder-Based Single-Cell RNA Sequencing Data Integration Algorithm Using Mutual Nearest...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• High-dimensional single-cell RNA sequencing (scRNA-seq) data assembly and downscaling are crucial for downstream analysis.
• Simultaneously eliminating batch effects and preserving cell topology in scRNA-seq datasets presents a significant challenge due to complex cell interrelationships.

Purpose of the Study:

• To introduce scGAMNN, a deep learning model utilizing a graph autoencoder architecture.
• To achieve simultaneous batch correction and topology-preserving dimensionality reduction for scRNA-seq data integration.

Main Methods:

• Developed scGAMNN, a deep learning model based on graph autoencoder.
• Applied scGAMNN to integrate scRNA-seq datasets, focusing on batch effect removal and topology preservation.
• Evaluated the low-dimensional integrated data for visualization, clustering, and trajectory inference.

Main Results:

• scGAMNN successfully performs batch correction and dimensionality reduction while maintaining the topology of cells within datasets.
• The integrated data from scGAMNN is suitable for various downstream analyses, including clustering and trajectory inference.
• Comparative analysis demonstrated that scGAMNN achieves comparable or superior performance in data integration, clustering, and trajectory conservation against five other methods.

Conclusions:

• scGAMNN offers a robust solution for integrating scRNA-seq data, addressing the dual challenge of batch correction and topology preservation.
• The model's ability to provide accurate low-dimensional representations facilitates reliable downstream biological interpretation.
• scGAMNN represents a significant advancement in scRNA-seq data analysis, improving the quality and utility of integrated datasets.