Adversarial dense graph convolutional networks for single-cell classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Single-cell transcriptomics aims to identify cell types and gene relationships.
• Data heterogeneity and noise in single-cell RNA sequencing (scRNA-seq) data present analysis challenges.

Purpose of the Study:

• To develop an advanced computational model for accurate single-cell classification.
• To address data heterogeneity and noise in scRNA-seq data analysis.

Main Methods:

• Proposed an adversarial dense graph convolutional network (HNNVAT).
• Incorporated dense connectivity and attention-based feature aggregation for enhanced feature learning.
• Utilized a feature reconstruction module to preserve original data features.
• Employed virtual adversarial training to improve model generalization and robustness.

Main Results:

• HNNVAT demonstrated superior performance compared to classical methods.
• Achieved higher classification accuracy on benchmark scRNA-seq datasets.
• The model effectively handles data heterogeneity and noise.

Conclusions:

• HNNVAT offers a robust and accurate approach for single-cell classification.
• The proposed architecture effectively integrates advanced deep learning techniques for scRNA-seq data analysis.
• The method has the potential to advance cell type identification and gene relationship studies.