Code accompanying our project on "Identifcation of phenotypic switching of colorectal tumour buds in response to tumour-microenvironment using spatially resolved transcriptomics", abstract available here.
This repo contains code for all analysis steps, from gene counts to pathway enrichment analyses and spatial deconvolution.
The data processing steps were handled by Nanostring's GeoMx NGS pipeline v2.3.3.10.
We built custom cell profile matrices from colorectal cancer single-cell data for the deconvolution tasks. The data is freely accessible from Pelka & al.'s article "Spatially organized multicellular immune hubs in human colorectal cancer".
Two matrices are available in ./Master_files: the "top" matrix groups together similar cell types, and the "mid" categories are a bit more detailed. Short descriptions for the cell type names are available here.
All R packages/dependencies can be installed at once using the renv lockfile in ./Scripts. renv allows us to build and share environment files in a similar way to conda. Once renv is installed and your R project is initiated, run renv::restore() to load the environment, making sure that renv.lock is in the project directory.
Note: This was built in R version 4.2.2 and although renv tracks the R version, it cannot change/update it. You might have to manually update some packages if your R version is different.
Quality control of the data processing results is done using script 0. The data exploration steps (PCA, GSVA) are covered in script 1. Details on the differential gene expression, GSEA and ORA can be found in script 2. The spatial deconvolution is in script 3.
The two remaining scripts were to build the custom profile matrices: ./Scripts/MK_CSTM_PRF_MTX.R
And to check for contamination in the epithelial compartment: ./Scripts/2B_TB_DCV.R