smoppix

smoppix unifies nonparametric tests for univariate and bivariate spatial patterns in single-molecule spatial omics data using the probabilistic index framework.

Key Features:

• Univariate spatial pattern tests: Tests for aggregation of transcripts and other univariate localization patterns across tissues and within cells.
• Bivariate spatial pattern tests: Tests for colocalization of transcript pairs and other bivariate localization patterns.
• Gradient detection: Tests for spatial gradients in molecule localization patterns.
• Vicinity tests: Tests for proximity of molecules to cell edge or cell centroid.
• Nested design support: Tests account for nested experimental designs across biological specimens.
• Automatic calibration: Tests are automatically calibrated to different biological specimens.
• Variance weighting: High-dimensionality of the data is exploited for variance weighting.
• Individual null distributions: Provides a meaningful background null distribution unique for every individual molecule.
• Exact permutation null distribution: Scales to high numbers of molecules and replicates through exact permutation.
• Segmentation-free analysis: Sidesteps segmentation, warping, edge correction and density estimation.

Scientific Applications:

• Transcript aggregation analysis: Identifies spatial aggregation patterns of individual transcripts in spatial omics data.
• Transcript colocalization analysis: Detects colocalization between pairs of transcripts at subcellular resolution.
• Cross-tissue spatial analysis: Performs spatial pattern tests across tissues as well as within cells.
• Multi-kingdom spatial omics: Applicable to datasets from diverse organisms across different kingdoms.
• Plant root spatial analysis: Validated experimentally on Selaginella moellendorfii roots.

Methodology:

smoppix is based on the probabilistic index, a nonparametric framework that unifies tests for multiple localization patterns. The method uses exact permutation null distributions and exploits high data dimensionality for variance weighting and individual molecule-specific background null distributions.