scVGATAE: A Variational Graph Attentional Autoencoder Model for Clustering Single-Cell RNA-seq Data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables the identification of cellular heterogeneity and developmental trajectories.
• Accurate cell subset identification is critical for scRNA-seq data analysis.
• Existing unsupervised clustering methods struggle with dropout, high dimensionality, noise, and computational time.

Purpose of the Study:

• To develop a novel unsupervised clustering method for scRNA-seq data.
• To improve the accuracy and efficiency of cell subpopulation identification.
• To address the limitations of existing clustering approaches.

Main Methods:

• Proposed scVGATAE (Single-cell Variational Graph Attention Autoencoder), a method integrating graph attention networks and variational autoencoders.
• Constructed a denoised cell graph to capture cell correlations.
• Employed adaptive training iterations and k-means clustering on learned low-dimensional representations.

Main Results:

• scVGATAE demonstrated superior performance compared to classical and state-of-the-art clustering methods.
• The method effectively handles noise and high dimensionality inherent in scRNA-seq data.
• Achieved accurate identification of cell subpopulations across nine public datasets.

Conclusions:

• scVGATAE offers a robust and efficient solution for unsupervised clustering of scRNA-seq data.
• The proposed method enhances the precision of cell subset identification.
• This advancement contributes to a deeper understanding of cellular heterogeneity and biological processes.