title: "Vignettes" author: "Dylan Cable" data: December 22, 2021

output: github_document

output: html_document: keep_md: yes pdf_document: default rmarkdown::html_vignette: keep_md: yes vignette: | %\VignetteIndexEntry{vignette-readme} %\VignetteEncoding{UTF-8} %\VignetteEngine{knitr::rmarkdown}

Vignettes: cell type identification and cell type-specific differential expression in spatial transcriptomics

Here, we will present examples and tutorials for how to run our computational methods for cell type identification (RCTD) and differential expression (C-SIDE) on spatial transcriptomics datasets. You may access RCTD and C-SIDE within our open-source R package here.

In total, we currently have 9 vignettes demonstrating the various applications of our software. These vignettes will be organized and referenced below, including applications to Slide-seq, Visium, and MERFISH data. In general, you can either view the html files (as linked to here) to view the vignette output, or you can run the raw R-markdown files on your own machine.

Cell type identification with RCTD

The best vignette for getting started with RCTD is spatial transcriptomics vignette.

RCTD can assign single cell types or cell type mixtures to spatial transcriptomics spots. RCTD has three modes: doublet mode, which assigns 1-2 cell types per spot and is recommended for technologies with high spatial resolution such as Slide-seq and MERFISH; full mode, which assigns any number of cell types per spot and is recommended for technologies with poor spatial resolution such as 100-micron resolution Visium; multi mode, an extension of doublet mode that can discover more than two cell types per spot as an alternative option to full mode. We demonstrate each mode in the following figures:

• Doublet mode: spatial transcriptomics vignette. Also, most other vignettes use doublet mode.
• Doublet mode on MERFISH: MERFISH nonparametric vignette.
• Full mode: full mode on Visium hippocampus
• Multi mode: multi mode on Visium hippocampus

Cell type-specific differential expression with C-SIDE

The best vignette for getting started with C-SIDE is differential expression vignette.

C-SIDE can detect differential expression (DE) along one or multiple user-defined axes, termed explanatory variables. Although the possibilities are not limited to what is presented here, we present the following examples:

• DE across two regions: C-SIDE across two regions in Slide-seq cerebellum.
• DE across more than two regions: Categorical C-SIDE on Visium hippocampus
• DE from cell-to-cell interactions: C-SIDE for cell-to-cell interactions.
• DE from interactions with pathology: C-SIDE for pathology interactions.
• Nonparametric smooth spatial patterns on MERFISH: MERFISH hypothalamus nonparametric vignette.

Batch processing of multiple experimental replicates + population-level DE inference

Finally, when multiple experimental replicates are available, RCTD and C-SIDE can be run in batch across replicates as shown in Population-level RCTD and C-SIDE. This approach also allows for population-level differential expression statistical inference.