A spatially informed matrix normal model for gene co-expression analysis in spatial transcriptomics studies.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatially resolved transcriptomics (SRT) advances gene expression profiling within tissue context.
• Gene co-expression analysis in SRT data is crucial for understanding gene function in the tissue microenvironment.
• Current methods struggle to integrate gene-gene interactions and spatial dependencies, limiting biological interpretation.

Purpose of the Study:

• To introduce spMOCA (SPatially informed Matrix-nOrmal model for gene Co-expression Analysis), a novel statistical framework.
• To infer gene co-expression networks by explicitly modeling spatial dependencies in SRT data.
• To improve the biological interpretability of gene co-expression in spatial transcriptomics.

Main Methods:

• Developed spMOCA, a statistical framework utilizing a matrix-normal model.
• Jointly modeled gene-gene and spatial covariance to disentangle intrinsic co-expression from spatial effects.
• Validated through extensive simulations and applications to nine diverse SRT datasets.

Main Results:

• spMOCA provides more accurate and unbiased gene-gene correlation estimates compared to existing methods.
• Consistently identified more experimentally validated transcription factor target genes across various SRT datasets.
• Uncovered tumor-related gene modules, prognostic markers, neurodegeneration-associated co-expression changes, and conserved cross-species pathways.

Conclusions:

• spMOCA effectively models spatial dependencies to reveal true gene co-expression relationships in SRT data.
• The method enhances biological discovery, identifying key genes and pathways in cancer, neurodegeneration, and cross-species comparisons.
• spMOCA represents a significant advancement for gene co-expression analysis in spatial transcriptomics.