Work

Spectrast: a multiscale spatial statistics toolkit for single-cell spatial transcriptomics

Year: 2026
Type: preprint
Abstract: Spatial transcriptomics resolves gene expression in tissue context, but the neighborhood statistics used to quantify spatial organization scale poorly to imaging-based datasets with hundreds-of-thousands of cells and thousands of genes. We recast the standard spatial-autocorrelation statistics as convolutions on a rasterized expression lattice and evaluate them with the fast Fourier transform, giving a single engine whose cost is governed by the neighborhood radius rather than the cell count. The engine computes each statistic across a continuous range of spatial scales from one shared transform, extends naturally to a differential test of spatial reorganization between tissue states, and runs unchanged on multicore CPUs or GPUs.
Institution: Dartmouth College
Language: English
Open Access status: green
OpenAlex ID: W7234554425
Cites: 43
Cited by:
Related to: 10