Single-cell RNA sequencing data analysis utilizing multi-type graph neural networks.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables cellular-level research but faces challenges in data analysis.
• Massive data, technical noise, and visualization difficulties hinder scRNA-seq data interpretation.
• Existing methods struggle with dimensionality reduction, denoising, and accurate cell clustering.

Purpose of the Study:

• To develop an advanced computational model for scRNA-seq data analysis.
• To improve dimensionality reduction, denoising, and cell clustering for scRNA-seq datasets.
• To enhance the understanding of cellular characteristics through robust data processing.

Main Methods:

• Proposed a novel single-cell data analysis model named scDMG.
• Integrated a zero-inflated negative binomial (ZINB) model with a denoising autoencoder (DAE) for dimensionality reduction and denoising.
• Employed multi-type graph neural networks for enhanced data preprocessing and feature learning, addressing dropout events.

Main Results:

• scDMG effectively performs dimensionality reduction and denoising on raw scRNA-seq data.
• The model demonstrates superior performance in resolving dropout events and enabling preliminary cell classification.
• Utilizing TSNE, PCA, and Louvain algorithms, scDMG achieved optimized dimensionality reduction and clustering.

Conclusions:

• scDMG outperforms existing scRNA-seq clustering algorithms in various performance metrics.
• The proposed model exhibits better scalability and shorter runtime compared to state-of-the-art methods.
• scDMG provides robust and efficient clustering results for diverse scRNA-seq datasets.