SpaMWGDA: Identifying spatial domains of spatial transcriptomes using multi-view weighted fusion graph convolutional...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics (ST) integrates gene expression with cellular spatial information.
• Deep learning (DL) methods have advanced spatial domain identification in ST.
• Existing DL methods have limitations in utilizing neighborhood information and integrating gene expression data.

Purpose of the Study:

• To address limitations in current DL-based ST analysis methods.
• To propose a novel DL model, SpaMWGDA, for enhanced spatial domain identification.
• To improve the integration of gene expression and spatial information for better tissue analysis.

Main Methods:

• Developed SpaMWGDA, a DL model using multi-view weighted fused graph convolutional network (GCN) and data augmentation.
• Modeled spatial information with diverse similarity metrics to capture comprehensive neighborhood data.
• Employed data augmentation and contrastive learning for key gene expression learning.
• Utilized a multi-view GCN encoder and view-level attention for adaptive feature integration.

Main Results:

• SpaMWGDA demonstrated superior performance in spatial domain identification compared to existing methods.
• The model achieved improved trajectory inference capabilities.
• SpaMWGDA effectively analyzed tissue structure and function, highlighting its analytical power.

Conclusions:

• SpaMWGDA offers a significant advancement in spatial transcriptomics analysis.
• The model's ability to integrate multi-view data and learn key features enhances biological insights.
• SpaMWGDA provides a powerful tool for understanding tissue organization and cellular interactions.