stDyer-image improves clustering analysis of spatially resolved transcriptomics and proteomics with morphological...

Area of Science:

• Computational biology
• Bioinformatics
• Deep learning for spatial omics

Background:

• Spatially resolved transcriptomics (SRT) and proteomics (SRP) provide gene and protein data in tissue context.
• Morphology images often accompany SRT/SRP data but are underutilized in analysis.
• Existing methods struggle to integrate image data effectively for clustering.

Purpose of the Study:

• To develop a novel deep learning framework, stDyer-image, for clustering SRT and SRP datasets.
• To leverage morphology images directly for enhanced clustering performance.
• To provide a versatile tool for large-scale spatial omics data analysis.

Main Methods:

• Developed an end-to-end deep learning framework, stDyer-image.
• Directly linked image features to cluster labels, inspired by pathologist image interpretation.
• Integrated morphology image analysis within the clustering pipeline.

Main Results:

• stDyer-image demonstrated superior clustering performance compared to state-of-the-art methods.
• The framework successfully integrated image features for improved spatial omics data clustering.
• Achieved high performance on large-scale datasets across diverse SRT/SRP technologies.

Conclusions:

• stDyer-image offers a powerful new approach for spatial omics data analysis.
• Directly utilizing image features significantly enhances clustering accuracy.
• The framework is versatile and applicable to various spatial omics datasets and technologies.