GSE179994_Tcells

Single-cell RNA-seq analysis workflow for T cell populations from GEO dataset GSE179994.

Project goal

This repository reproduces preprocessing and clustering of T cell scRNA-seq data and creates post-treatment CD4 and CD8 subsets for downstream comparative analysis between responders and non-responders to immunotherapy.

Repository structure

• README.md: project overview and run instructions.
• Scripts/SingleCellFunctions.R: reusable helper functions.
• Scripts/scRNAseq.R: main analysis pipeline.
• Scripts/run_analysis.R: one-command entrypoint.
• Scripts/proportion_analysis.ipynb: downstream proportion analysis notebook.

During execution, outputs are written to Figures/ and Results/.

Requirements

• R 4.3.0 or newer.
• Seurat, dplyr, scuttle, ggplot2.
• Optional for AnnData export: SeuratDisk.

Install packages if needed:

install.packages(c("Seurat", "dplyr", "scuttle", "ggplot2"))

Install SeuratDisk if you want .h5ad outputs:

install.packages("remotes")
remotes::install_github("mojaveazure/seurat-disk")

Input data

Expected input files in repository root or GSE179994_RAW/:

• GSE179994_all.Tcell.rawCounts.rds
• GSE179994_Tcell.metadata.tsv

Required metadata column:

• cellid

How to run

From repository root:

Rscript Scripts/run_analysis.R

Pipeline steps:

1. Load counts and metadata.
2. Build and preprocess Seurat object.
3. Add treatment labels (Pre/Post from sample name).
4. Run clustering and UMAP.
5. Subset post-treatment CD4 and CD8 populations.
6. Flag low-quality cells using 2 MAD outlier thresholds.
7. Harmonize metadata fields for AnnData compatibility (cluster, Response when present).
8. Save figures, RDS outputs, and export CD4/CD8 post-treatment subsets to h5ad (if SeuratDisk is installed).

Output files

• Figures/UMAP_Tcell.pdf
• Figures/UMAP_post_CD4_Tcell.pdf
• Figures/UMAP_post_CD8_Tcell.pdf
• Results/Seurat_Reanalyzed.rds
• Results/GSE179994_post_CD4_reanalyzed.rds
• Results/GSE179994_post_CD8_reanalyzed.rds
• Results/CD4_filtered_post.h5ad (if SeuratDisk installed)
• Results/CD8_filtered_post.h5ad (if SeuratDisk installed)

Methods summary

• CreateSeuratObject filters: min.cells = 3, min.features = 200.
• Cell-level QC: nFeature_RNA > 600, nFeature_RNA < 25000, nCount_RNA > 600.
• Seurat workflow: NormalizeData, FindVariableFeatures, ScaleData, PCA, neighbors, clustering, UMAP.
• CD4/CD8 post-treatment QC with scuttle::isOutlier using nmads = 2.

References

• Seurat v4 tutorial: https://satijalab.org/seurat/archive/v4.3/pbmc3k_tutorial
• Original publication: https://www.nature.com/articles/s43018-021-00292-8#Sec9
• Outlier-based QC rationale: https://genomebiology.biomedcentral.com/articles/10.1186/s13059-020-02136-7
• Differential abundance context: https://www.nature.com/articles/s41598-024-66381-7
• scanpro: https://github.com/loosolab/scanpro