Single-cell classification using graph convolutional networks

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNAseq) is crucial for understanding cellular processes.
• Cell type identification is a priority in scRNAseq data analysis.
• Gene interaction networks enhance cell classification accuracy.

Purpose of the Study:

• To develop a novel deep learning model for cell classification.
• To integrate gene expression data with gene interaction networks.
• To improve the accuracy of cell type identification.

Main Methods:

• Proposed a multimodal end-to-end deep learning model, sigGCN.
• Combined a graph convolutional network (GCN) with a neural network.
• Evaluated performance using standard classification metrics on within-dataset and cross-dataset tasks.

Main Results:

• sigGCN demonstrated superior performance compared to existing cell classification tools.
• The model achieved higher classification accuracy and F1 scores.
• Compared performance against traditional machine learning methods.

Conclusions:

• Integrating gene interaction networks with gene expression data improves cell classification.
• GCN methodologies effectively extract features for enhanced cell identification.
• sigGCN offers a powerful approach for cell type classification in scRNAseq data.