BetaReg_ST

This repository contains software for processing and analysing Visium HD data from 8 wpc pancreatic samples.
It is written in R and uses the targets package to build a fully reproducible analysis pipeline.

The pipeline definitions can be found in the following files:

• R/tar_main.R
• R/tar_8um.R
• R/tar_2um.R

Running the targets Pipeline

Some stages of the pipeline are memory-intensive and require long runtimes. These steps should be executed on a high-memory cluster node with minor adjustments to the workflow.

Set up the R environment

Create and restore the analysis environment using renv:

install.packages("renv")
renv::restore()

Prepare input data for the external script

Run the following command to generate the input files:

targets::tar_make(tidyselect::starts_with("sav_external_cor"))

This will produce files named ext/seu_?.rds, where ? is the sample number (e.g., 1 and 2 in our case).

Run the correlation script on the cluster

Copy the .rds files and the script scripts/cscore_correlation_all.R to the cluster.
Run the script for each file:

cscore_correlation_all.R seu_1.rds cor_1.parquet
cscore_correlation_all.R seu_2.rds cor_2.parquet

Notes:

• It is recommended to submit these jobs via qsub.
• Request at least 100 GB of RAM.
• On our cluster, these jobs took 12–16 hours to complete.

Retrieve and link the results

When the jobs finish, copy the generated .parquet files back to your laptop into the shiny/data directory (for use in the Shiny apps). Also, create symbolic links (rather than duplicating these large files) in the ext directory.

Complete the pipeline

Resume the pipeline:

targets::tar_make()

Generate the report and run Shiny apps

Render the Quarto report:

quarto::quarto_render("doc/report.qmd")

Launch the Shiny applications:

shiny::runApp("shiny/de")
shiny::runApp("shiny/gsea")
shiny::runApp("shiny/gene_explorer")