Not finished yet. It has not been run end to end. New to single-cell? Start with the beginner tutorial instead.
A full 10x single-nucleus workflow on a public human kidney dataset (GSE131882, 3 control vs 3 early diabetic nephropathy; Wilson et al., PNAS 2019). Each step states the decision, the default, and when to deviate.
Cell Ranger outs/ per sample, containing both raw_feature_bc_matrix/ and filtered_feature_bc_matrix/. The raw matrix is required for ambient RNA correction, which is why this starts from FASTQs (SRA) and not from processed GEO tables. Paths go in samplesheet.csv.
GEO only has zUMIs count matrices as RDS (already barcode-filtered), so there is no raw droplet matrix to download. FASTQs are in SRA under SRP199727 (BioProject PRJNA545232), samples GSM3823939 to GSM3823944. Libraries are 10x Chromium Single Cell 5’.
Suggested route: nf-core/fetchngs for SRP199727, then nf-core/scrnaseq with --aligner cellranger on GRCh38. Introns are counted by default in Cell Ranger 7 and later. Point samplesheet.csv at each sample’s outs/ folder.
./run.sh check # package versions inside the container
./run.sh render # renders tutorial.html
ENGINE=apptainer ./run.sh render # on HPC (run from this folder)
Container: ghcr.io/armilner/r-bioinformatics-2026
After the first render, review results/annotation_proposed.csv against the marker dot plot, edit the celltype column, save it as annotation_approved.csv in this folder, and render again.
results/cluster_markers.csv, results/pseudobulk_DE.csv, results/gsea.csv, results/kidney_snrna_annotated.rds, tutorial.html