ScGSLC: An unsupervised graph similarity learning framework for single-cell RNA-seq data clustering.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate cell clustering is crucial for cell type identification in single-cell RNA sequencing (scRNA-seq) data.
• High dimensionality and sparsity in scRNA-seq data challenge traditional clustering methods.
• Understanding cellular function relies on genetic networks, where deep learning excels at representation learning.

Purpose of the Study:

• To propose a novel scRNA-seq clustering framework, ScGSLC, leveraging graph similarity learning.
• To enhance the accuracy of cell clustering by integrating diverse biological data sources.
• To improve cell type identification through advanced computational methods.

Main Methods:

• ScGSLC integrates scRNA-seq data with protein-protein interaction networks into a graph structure.
• Graph convolution networks are employed for graph embedding and similarity calculation.
• Unsupervised clustering is performed based on the learned graph similarities.

Main Results:

• ScGSLC demonstrated superior performance in unsupervised clustering across nine public scRNA-seq datasets.
• The proposed method effectively captures cell similarities by integrating network information.
• Results indicate ScGSLC outperforms existing state-of-the-art clustering techniques.

Conclusions:

• ScGSLC offers a robust and accurate framework for scRNA-seq data clustering.
• The integration of genetic networks and deep learning provides significant advantages for cell analysis.
• This approach advances the field of computational biology and cell type identification.