DenMark - Cell-Density Correlated Genes from Spatial Transcriptomics

Quantifies the dependence between local cell density and gene expression in single-cell resolution spatial transcriptomics data, such as data from the Xenium and MERFISH platforms. Cell locations form a point pattern and gene counts are its marks. After aggregation on a regular grid, cell counts and summed expression follow Poisson models driven by two coregionalized latent Gaussian fields with Matern covariance, approximated by Hilbert space basis functions; the correlation between the two fields measures the density-expression dependence. Models are fitted with 'Stan' through 'cmdstanr'. Functions cover the tissue boundary (an alpha shape of the cells), gridding, model fitting, posterior summaries, latent-field decomposition, spatial correlation ranges, and gene scans that call density-correlated genes.