SpatialESD: Spatial Ensemble Domain Detection in Spatial Transcriptomics.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics (ST) enables gene expression measurement within tissue context.
• Spatial domain detection in ST data is challenging due to data complexity and variable clustering method performance.

Purpose of the Study:

• To develop an ensemble method, SpatialESD, for improved spatial domain detection in ST data.
• To enhance the robustness and accuracy of spatial domain identification by integrating diverse clustering approaches.

Main Methods:

• SpatialESD integrates results from multiple spatial domain detection algorithms.
• The method captures direct co-occurrence and multiscale indirect relationships between clusters.
• Evaluation performed on simulated and real-world ST datasets (human brain, breast, ovarian cancer).

Main Results:

• SpatialESD consistently outperformed individual methods and the existing EnSDD ensemble.
• Demonstrated superior clustering accuracy and stability across diverse datasets.
• Facilitated downstream analyses including differential gene expression, trajectory, and cell-cell interaction analysis.

Conclusions:

• SpatialESD offers a reliable and effective solution for spatial domain detection in ST data.
• The method enhances understanding of tissue organization and disease mechanisms through improved spatial analysis.
• Enables more robust downstream biological discovery from spatial transcriptomics.