Deep learning in integrating spatial transcriptomics with other modalities.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics (ST) enables studying gene expression within tissue context.
• ST platforms often integrate histology, chromatin images, and single-cell RNA sequencing (scRNA-seq) data.
• Integrating multimodal data enhances understanding of tissue molecular landscapes.

Purpose of the Study:

• To systematically review deep learning (DL) methods for integrating ST with other data modalities.
• To categorize DL techniques by integrated modality and task.
• To identify challenges and future directions in spatial multi-omics integration.

Main Methods:

• Systematic literature review of DL applications in ST integration.
• Delineation of DL techniques and key integration tasks.
• Categorization of methods based on integrated modalities (e.g., imaging, scRNA-seq) and tasks (e.g., data fusion, imputation).

Main Results:

• Identified and categorized various DL approaches for ST integration.
• Summarized common integration strategies employed by these methods.
• Highlighted the growing trend of spatial multi-omics data integration.

Conclusions:

• Deep learning is crucial for integrating diverse spatial omics data.
• Standardized methods and benchmarks are needed for robust multimodal analysis.
• Future research should focus on developing advanced DL models for comprehensive spatial multi-omics interpretation.