Addressing the mean-variance relationship in spatially resolved transcriptomics data with spoon.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Identifying spatially variable genes (SVGs) is crucial for analyzing spatial transcriptomics data.
• Current methods for SVG ranking may be affected by the mean-variance relationship, a technical bias observed in RNA-sequencing data.
• This bias can lead to inaccurate prioritization of genes based on expression levels and variance.

Purpose of the Study:

• To demonstrate the presence of the mean-variance relationship in spatial transcriptomics data.
• To introduce spoon, a novel statistical framework designed to mitigate this bias.
• To improve the accuracy of identifying and prioritizing spatially variable genes.

Main Methods:

• Demonstration of the mean-variance relationship in spatial transcriptomics datasets.
• Development of spoon, a statistical framework employing Empirical Bayes techniques.
• Validation of spoon using simulated and real-world spatial transcriptomics data.

Main Results:

• Confirmation of the mean-variance relationship in spatial transcriptomics.
• spoon effectively removes the technical bias, leading to more accurate SVG identification.
• Improved prioritization of SVGs compared to existing methods.

Conclusions:

• The mean-variance relationship poses a challenge for accurate SVG identification in spatial transcriptomics.
• spoon provides a robust solution for bias correction, enhancing the reliability of spatial gene expression analysis.
• The spoon software implementation facilitates broader application of this improved methodology.