Contrastive graph regularized non-negative matrix factorization for domain identification of spatial transcriptomics.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics provides gene expression data with spatial resolution, but high dimensionality hinders spatial domain identification.
• Existing deep learning methods offer feature extraction but lack interpretability, necessitating dimensionality reduction techniques that preserve spatial and biological relevance.

Purpose of the Study:

• To propose a novel contrastive graph-regularized non-negative matrix factorization (CGNMF) model for interpretable dimensionality reduction in spatial transcriptomics.
• To enhance feature representation and spatial structure preservation for improved spatial domain identification.

Main Methods:

• Developed a CGNMF model integrating graph regularization with a self-supervised contrastive learning framework.
• Constructed positive and negative sample pairs using gene expression similarity and spatial proximity.
• Incorporated contrastive learning into graph-regularized non-negative matrix factorization to guide factorization towards biologically and spatially coherent dimensions.

Main Results:

• CGNMF consistently outperformed seven existing spatial domain identification methods across three public datasets based on clustering metrics.
• The model successfully identified biologically relevant functional regions missed by current approaches.
• Demonstrated improved interpretability and automatic delineation of spatial domains.

Conclusions:

• CGNMF offers a robust and interpretable approach for dimensionality reduction in spatial transcriptomics.
• The method enhances the identification of spatial domains and biological insights from spatial transcriptomics data.
• CGNMF represents a significant advancement in analyzing high-dimensional spatial transcriptomics data.