scDFC: A deep fusion clustering method for single-cell RNA-seq data.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Clustering is crucial for analyzing tumor heterogeneity in single-cell RNA sequencing (scRNA-seq) data.
• Traditional clustering methods struggle with high-dimensional scRNA-seq data.
• Deep clustering methods show promise but often neglect integrating cell attribute and structure information simultaneously.

Purpose of the Study:

• To develop a novel deep fusion clustering model for scRNA-seq data.
• To effectively integrate both attribute and structure information for enhanced clustering performance.
• To improve the investigation of cell subpopulations and tumor microenvironments.

Main Methods:

• Proposed a single-cell deep fusion clustering model with two modules: attributed feature clustering and structure-attention feature clustering.
• Utilized two autoencoders to process diverse feature types.
• Integrated attribute, structure, and attention information for comprehensive analysis.

Main Results:

• The proposed model demonstrates the validity and efficiency of fusing attribute, structure, and attention information.
• Experimental results confirm the model's effectiveness on scRNA-seq data.
• The approach facilitates a more nuanced understanding of cellular heterogeneity.

Conclusions:

• The novel deep fusion clustering model offers a powerful tool for scRNA-seq data analysis.
• This method enhances the ability to identify cell subpopulations and study tumor microenvironments.
• The open-source implementation facilitates further research and application.