Reg2ST: recognizing potential patterns from gene expression for spatial transcriptomics prediction.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Spatial transcriptomics provides gene expression data with spatial context, crucial for various biological studies.
• Current sequencing methods are expensive and slow, hindering research progress.
• Deep learning shows promise for spatial transcriptomics prediction but faces limitations.

Purpose of the Study:

• To develop an advanced deep learning model, Reg2ST, for accurate spatial transcriptomics prediction.
• To address limitations in existing deep learning methods for integrating spatial transcriptomics and histology data.
• To improve the efficiency and cost-effectiveness of spatial transcriptomics analysis.

Main Methods:

• Reg2ST employs contrastive learning to align spatial transcriptomics and histology data.
• It predicts gene expression patterns using histology image features.
• The model captures spot relationships uniquely, moving beyond K-Nearest-Neighbors.

Main Results:

• Reg2ST demonstrates superior performance in spatial gene expression prediction.
• Evaluations using human breast cancer and cutaneous squamous cell carcinoma datasets confirm its effectiveness.
• The model shows improved computational efficiency compared to existing methods.

Conclusions:

• Reg2ST offers a powerful and efficient approach for spatial transcriptomics prediction.
• The model's ability to integrate histology and gene expression data opens new avenues for biological discovery.
• Reg2ST advances the field by providing a more accessible tool for spatial transcriptomics research.