Multi-modal domain adaptation for revealing spatial functional landscape from spatially resolved transcriptomics.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Spatially resolved transcriptomics (SRT) offers insights into gene expression within tissue microenvironments.
• Integrating multimodal SRT data (gene expression, histology, spatial location) is challenging due to expression sparsity.
• Existing methods struggle with comprehensive analysis of complex spatial transcriptomic datasets.

Purpose of the Study:

• To introduce stMDA, a novel unsupervised domain adaptation method for integrating multimodal spatial transcriptomics data.
• To reveal the spatial functional landscape by combining gene expression with other modalities.
• To improve spatial clustering and variation analysis in SRT datasets.

Main Methods:

• stMDA employs neural networks to learn modality-specific representations from spatial multimodal data.
• It aligns spatial distributions across these representations for effective data integration.
• The method integrates global and spatially local information for enhanced clustering consistency.

Main Results:

• stMDA demonstrates superior performance in identifying spatial domains across various platforms and species compared to existing methods.
• The approach successfully identifies spatially variable genes with significant prognostic value in cancer tissues.
• Results highlight stMDA's capability in handling expression sparsity for robust spatial analysis.

Conclusions:

• stMDA provides a powerful and flexible framework for multimodal data integration in spatial transcriptomics.
• This new tool advances the analysis of SRT datasets, deepening our understanding of biological systems.
• stMDA facilitates the exploration of intricate spatial relationships and their impact on biological functions and disease.