SpaceX: gene co-expression network estimation for spatial transcriptomics

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatially resolved transcriptome analysis reveals cellular interactions and transcriptional regulation.
• Gene-gene co-expression patterns within distinct tissue locations are crucial for understanding spatial co-regulatory networks.
• Existing methods often focus on single gene analysis, limiting the discovery of complex spatial interactions.

Purpose of the Study:

• To develop a statistical framework for detecting gene co-expression patterns in spatially structured tissues.
• To enhance the capabilities of spatial transcriptomics technologies for biological discovery.
• To identify shared and cluster-specific co-expression networks across different cell types or tissue domains.

Main Methods:

• Development of SpaceX (spatially dependent gene co-expression network), a Bayesian methodology.
• Utilizing an over-dispersed spatial Poisson model combined with a high-dimensional factor model for efficiency.
• Incorporating dimension reduction techniques for computational performance.

Main Results:

• SpaceX accurately estimates co-expression networks and structures by accounting for spatial correlation and noise.
• Analysis of mouse hypothalamus data identified hub genes related to cognitive abilities.
• Analysis of human breast cancer data revealed cancer-associated genes, including the collagen family, in tumor regions.

Conclusions:

• SpaceX provides a robust statistical framework for analyzing spatial transcriptomics data.
• The method effectively identifies biologically relevant gene co-expression patterns in complex tissues.
• SpaceX advances the potential of spatial transcriptomics for uncovering novel biological insights.