IMMUcan

Processing scripts for scRNA-seq database

Install instructions

• Follow install instructions for sceasy (https://github.com/cellgeni/sceasy)
• Get CHETAH_reference_updatedAnnotation.RData from IMMUcan teams channel
• Install following R packages

install.packages(c("Seurat", "tidyverse", "readxl", "patchwork", "devtools", "data.table", "BiocManager", "remotes", "WriteXLS", "pheatmap", "plyr"))
BiocManager::install(c("CHETAH", "SingleCellExperiment"))
devtools::install_github("mahmoudibrahim/genesorteR") 
devtools::install_github("immunogenomics/harmony")
remotes::install_github("mojaveazure/seurat-disk")

Clone this repository

git clone https://github.com/soumelis-lab/IMMUcan.git

Before starting

Change the paths to files provided in the script

• cellMarker_path = PATH to TME_markerGenes.xlsx
• chetahClassifier_path = PATH to CHETAH_reference_updatedAnnotation.RData
• cellOntology_path = PATH to cell_ontology.xlsx

Run scProcessor

• Start from a seurat object
• scProcessor creates two folders where it stores files: temp/ and out/
• FILL IN THE SQUARED BRACKETS

1. Check Seurat object

Input:

• [SEURAT]: path to seurat object

Rscript check_seurat.R [SEURAT] 

Output:

• Prints cell_id, gene_id
• Checks if metadata is correctly linked
• Checks normalization: raw or normalized (stores it in data.rds)
• Prints meta.data slot

2. Select desired QC thresholds and batch variable

Input:

• [BATCH]: fill in part or full name of batch variable(s) e.g. "patient|sample|plate"

Rscript scProcessor_test.R [BATCH]

Output:

• QC plot and elbowplot in temp/

3. scProcessor_1

Bayer only: bash scProcessor_1.sh with slurm if on an HPC (adapt vars in bash file)

Input:

• [BATCH]: FULL name of the batch column in the metadata (often patient or sample), if no integration is desired use "none"
• [FEATURES]: minimum amount of detected genes per cell allowed
• [MITO]: maximum percentage of mitochondrial reads per cell allowed
• [PCA]: number of PCA dimensions to use for downstream processing

Rscript scProcessor_1.R [BATCH] [FEATURES] [MITO] [PCA]

Steps:

• QC
• Batch integration
• Dimensionality reduction + clustering
• Supervised classification

Output:

• Plots on QC, integration and supervised classification in out/
• Seurat clustering, markers gene plots and comparison seurat/supervised in temp/
• annotation.xls in out/

4. Annotate clusters

• Check plots in temp/:
  • marker gene plots
  • dotplot
• In out/annotation.xls, fill in cell types as defined in cell_ontology.xlsx in the abbreviation column

5. scProcessor_2

Bayer only: bash scProcessor_2.sh with slurm if on an HPC (adapt vars in bash file)

Rscript scProcessor_2.R

Steps:

• Links annotation to seurat_clusters
• Includes cell ontology
• Differential expression
• Creates output files for SIB scRNAseq interface

Output:

• AverageExpression matrices and DE_results per annotation level in out/
• geneIndex.tsv
• Metadata.tsv in out/
• cellCount.tsv in out/
• cellxgene.h5ad in out/

6. Create checksum file to send to SIB

on the terminal

cp out [DATASET_ID]_-_`md5sum /etc/group | awk -F" " '{print $1}'`
zip -r [DATASET_ID_-_checksum].zip [DATASET_ID_-_checksum]

Login to SIB through sftp and transfer