A Fusion Learning Model Based on Deep Learning for Single-Cell RNA Sequencing Data Clustering.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables cellular-level biological insights.
• Unsupervised clustering is crucial for identifying distinct cell types in scRNA-seq data.
• Existing clustering algorithms often overlook the integration of both superficial and deep data characteristics.

Purpose of the Study:

• To develop a deep learning-based fusion framework, scGASI, for enhanced scRNA-seq data clustering.
• To effectively combine diverse feature sets by integrating data affinity recovery and deep feature embedding.
• To improve the accuracy of cell type identification through a novel fusion learning approach.

Main Methods:

• Constructed a fusion learning framework (scGASI) utilizing deep learning.
• Integrated data affinity recovery and deep feature embedding for similarity matrix learning.
• Employed a graph autoencoder for low-dimensional latent representation and a self-expression-based fusion model to merge data features.

Main Results:

• scGASI successfully integrates surface and deep information from scRNA-seq data.
• The framework learns comprehensive similarity matrices for individual and all feature sets.
• Extensive validation shows scGASI surpasses widely used clustering methods in accuracy.

Conclusions:

• scGASI provides a robust method for single-cell type discrimination using scRNA-seq data.
• The fusion learning approach effectively leverages multi-faceted data information.
• scGASI demonstrates superior performance in clustering accuracy compared to existing techniques.