Skip to contents

Case studies

Source: vignettes/articles/case-studies.Rmd
case-studies.Rmd

The scripts under CodeInPaper/ reproduce the analyses of the DenMark manuscript. Release v1.0.0 will be the version that accompanies the manuscript, and later releases may change the results, so run the scripts from a checkout of that release once it is tagged.

Each script begins with a CONFIGURE ME block that sets DATA_DIR, OUTPUT_DIR and REPO_ROOT for your environment. Each subfolder has its own README.md, and CodeInPaper/README.md maps the scripts to the figures they produce, gives the order in which they run and lists any inputs not yet in the repository. That table is the authoritative one; the overview below follows the Results of the manuscript.

Subfolder Case study Results subsection Dataset access
Simulation/ False and true positive rates of DCG calls on simulated genes of known status Validation of DenMark on simulated data Simulated by the scripts
MERFISH-MouseBrain/ Astrocytes in Vizgen MERSCOPE mouse brain sections: section S3R1 in detail and nine sections for reproducibility Density-correlated genes in mouse-brain astrocytes complement cell-type markers and are reproducible across sections Vizgen public data release (Google Cloud Storage)
Xenium-BreastCancer/ Three regions of interest (invasive tumor, DCIS-1, DCIS-2) and whole-section fits of DCIS, invasive tumor and CD8+ T cells in a 10x Xenium breast tumor Density-correlated genes reveal the spatial antagonism of tumor and immune compartments in breast cancer 10x Genomics
OvarianCancer-VisiumHD-Xenium5K/ Density correlations of tumor-cell programs on adjacent Visium HD and Xenium Prime 5K sections of one ovarian tumor DenMark recovers program-level density correlations that transfer across platforms 10x Genomics
Visium-ColorectalCancer/ A 10x Visium CytAssist colorectal cancer section with nuclei counted from the H&E image, and Visium-like spots built from the Xenium breast section DenMark extends to spot-level Visium data and recovers single-cell density correlations after aggregation 10x Genomics

The mouse-brain sections and the ovarian Visium HD and Xenium sections are gridded over an alpha-shape tissue domain, which the packages now provide (denmark_alpha_shape() and boundary =). Two other features of these analyses arrive in the packages in v1.0.0: the total transcript count as the expression offset (Visium HD and Visium) and the pooling of Visium spots into super-grids. The scripts are kept as they were run and fit their models directly, with the Stan files bundled in inst/stan/ (the variable-area and total-UMI versions of the model) or kept next to the script. Three models that some scripts call are not yet in the repository; CodeInPaper/README.md lists them under Missing files.

Data sources