Deep structural clustering for single-cell RNA-seq data jointly through autoencoder and graph neural network.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables the study of cellular heterogeneity.
• Unsupervised clustering is crucial for identifying cell types in scRNA-seq data.
• Challenges in scRNA-seq analysis include noise, high dimensionality, and data sparsity.

Purpose of the Study:

• To develop a novel deep structural clustering method for scRNA-seq data.
• To improve the accuracy and scalability of cell type identification.
• To address the computational challenges associated with scRNA-seq data analysis.

Main Methods:

• Proposed scDSC method integrating structural information into deep clustering.
• Utilized a Zero-Inflated Negative Binomial (ZINB) model-based autoencoder for data representation.
• Incorporated a graph neural network (GNN) module to capture cell structural information.
• Employed a mutual supervised strategy to unify autoencoder and GNN modules for clustering.

Main Results:

• scDSC demonstrated superior performance across six real scRNA-seq datasets.
• The method showed significant improvements in clustering accuracy.
• scDSC exhibited enhanced scalability compared to existing state-of-the-art methods.

Conclusions:

• scDSC offers an effective solution for clustering scRNA-seq data.
• The integration of structural information and deep learning enhances cell type identification.
• The proposed method advances the analysis of complex single-cell data.