DenMark

DenMark (Density-dependent Marked Point process framework) is a model-based statistical framework to quantify how gene expression varies with local cell density and to identify density-correlated genes (DCGs). It is designed for single-cell resolution spatial transcriptomics data such as MERFISH, Xenium and SeqFISH, where cell location and gene expression at the single-cell resolution are provided.

---

DenMark Workflow

Implemented with a density-dependent marked point process, as well as comparing to the one with an independent marked point process, DenMark enables downstream analyses such as:

• jointly quantify the spatial heterogeneity of the cell locations and a typical gene expression (candidate gene);
• quantify the correlation between cell density and gene expression
• identify the DCGs in the provided single-cell resolution spatial transcriptomics dataset

This repository contains the reference R implementation used in the DenMark manuscript.

---

Repository layout

• DenMark/
  Core R implementation of DenMark (grid discretization and main modeling functions).

• CodeInPaper/
  Scripts used to generate the figures and results in the manuscript (simulation study, MERFISH mouse brain data, and Xenium breast cancer data).

• Images/
  Images in the paper and this repo.

• Tutorial_DenMark.rmd / Tutorial_DenMark.pdf / Tutorial_DenMark.html

  An RMarkdown tutorial that walks through two single-cell resolution datasets (MERFISH mouse brain and Xenium breast cancer data).

• LICENSE
  License for using and modifying this code.

---

A Quick Start

Please refer to the tutorial file Tutorial_DenMark.rmd.

---

Reproducing results from the manuscript

The scripts in CodeInPaper/ (to be documented) correspond to the main analyses:

• Simulation study
  Evaluates two approximation performances (grid-based approach vs. the actual marked point process; HSGP vs. exact GP).

• MERFISH mouse brain data Quantify the spatial heterogeneity in cell locations and gene expression, and candidate gene expression correlation to cell density. Identification of DCGs is also provided.

• Xenium breast cancer data Quantify the spatial heterogeneity in cell locations and gene expression, and candidate gene expression correlation to cell density. Identification of DCGs is also provided.

Data sources:

• Mouse brain MERFISH: https://console.cloud.google.com/storage/browser/public-datasets-vizgen-merfish ;

• Human breast cancer Xenium: https://www.10xgenomics.com/products/xenium-in-situ/preview-dataset-human-breast .