spoon

spoon addresses the mean-variance relationship in spatially resolved transcriptomics data by generating precision weights for individual observations using Empirical Bayes techniques, enabling more accurate prioritization of spatially variable genes.

Key Features:

• Mean-variance bias correction: spoon removes technical bias introduced by log-transformation that violates the mean-variance relationship of gene counts, where highly expressed genes have higher variance in counts but lower variance after log-transformation.
• Precision weight generation: Empirical Bayes techniques generate precision weights for individual observations in spatially resolved transcriptomics data.
• Data rescaling: Generated weights rescale the data and covariates for use as input in spatially variable gene detection tools.
• Ranking by effect size: The framework enables ranking of spatially variable genes based on effect sizes such as the proportion of spatial variance rather than solely P-values.

Scientific Applications:

• Spatially variable gene detection: spoon provides rescaled data and covariates as input for existing spatially variable gene detection tools to improve their accuracy.
• Spatial transcriptomics analysis: The framework prioritizes genes that vary in 2D space in spatially resolved transcriptomics datasets.
• Performance benchmarking: spoon has been validated in both simulated and real spatially resolved transcriptomics data.

Methodology:

spoon uses Empirical Bayes techniques to generate precision weights for individual observations, which are then used to rescale the data and covariates.