A reproducible, config-driven R pipeline for TCGA transcriptomic analysis — from raw counts to differential expression, functional enrichment, and survival outcomes.

---

Overview

MINT v1.0.0 takes a TCGA project (e.g. TCGA-BRCA, TCGA-LUSC) from raw expression data through a full differential expression and survival analysis workflow, producing a self-contained HTML report at the end. It was built around a single BRCA analysis and generalized so the same pipeline can be pointed at other TCGA cohorts by changing a config file, not the code.

Core capabilities:

• Automated TCGA RNA-seq counts + clinical data acquisition and caching (via TCGAbiolinks)
• Automatic Tumor/Normal label inference and low-count gene filtering, with a warning if too few Normal samples are available
• Differential expression analysis (DESeq2), with PCA, dispersion, volcano, MA plot, and top-DEG heatmap
• Functional enrichment: GO (Biological Process), KEGG, and Reactome over-representation analysis on up/down DEGs
• Gene-set enrichment analysis (GSEA) against an MSigDB collection
• Survival analysis: Kaplan–Meier curves plus univariate (and multivariate, where clinical covariates allow) Cox proportional hazards, stratified by a gene of interest
• A single auto-rendered HTML report (report.Rmd) embedding every plot and table, opened automatically on completion
• Checkpointed, segment-by-segment execution — rerun one stage without repeating the rest

Workflow

flowchart TD
    A[TCGA Project ID<br/>e.g. TCGA-BRCA] --> B[01: Download Data<br/>TCGAbiolinks RNA-seq + clinical, cached]
    B --> C[02: Preprocess Data<br/>counts matrix, Tumor/Normal labels,<br/>gene ID↔symbol map, low-count filtering]
    C --> D[03: DESeq2 Analysis<br/>differential expression, PCA,<br/>volcano, MA plot, top-DEG heatmap]
    D --> E[04: Functional Enrichment<br/>GO, KEGG, Reactome ORA]
    D --> F[05: GSEA<br/>full ranked list vs MSigDB collection]
    D --> G[06: Survival Analysis<br/>Kaplan–Meier + Cox regression,<br/>stratified by gene of interest]
    E --> H[07: Generate Report<br/>renders report.Rmd to HTML]
    F --> H
    G --> H
    H --> I[Final Output<br/>HTML report, opened automatically]

    style A fill:#eef7f0,stroke:#3a8f5f
    style I fill:#eef7f0,stroke:#3a8f5f

Loading

Each numbered stage lives in its own script under R/, run in sequence by run_pipeline.R. Shared helper functions live in utils.R. Intermediate objects are cached, so any stage can be re-run independently after a fix without repeating earlier steps.

Project structure

MINT/
├── run_pipeline.R              # driver — sources R/01 through R/07
├── config.R                    # all user-editable settings (project ID, cutoffs, gene of interest)
├── report.Rmd                  # R Markdown template rendered by 07
├── data/                      # all data files (creates on run automatically)
├── R/
│   ├── 01_download_data.R        # TCGAbiolinks download + cache (RNA-seq + clinical)
│   ├── 02_preprocess_data.R      # counts matrix, Tumor/Normal labels, gene ID↔symbol map, filtering
│   ├── 03_deseq2_analysis.R      # DESeq2 DE, PCA, dispersion, volcano, MA plot, top-DEG heatmap
│   ├── 04_functional_enrichment.R # GO (BP), KEGG, Reactome ORA on up/down DEGs
│   ├── 05_gsea_analysis.R        # GSEA on the full ranked gene list vs an MSigDB collection
│   ├── 06_survival_analysis.R # Kaplan–Meier + Cox PH, stratified by gene of interest
│   ├── 07_generate_report.R      # renders report.Rmd to HTML, opens it automatically
│   └── utils.R                     # shared helper functions used across stages
├── assets/
│   └── mint_logo.png
├── results/                     # figures, tables, and the rendered report land here (creates on run automatically)
└── README.md

Requirements

• R (≥ 4.2 recommended)
• Bioconductor: DESeq2, TCGAbiolinks, clusterProfiler, ReactomePA, EnhancedVolcano, org.Hs.eg.db
• CRAN: survival, survminer, msigdbr, ggplot2, dplyr, rmarkdown

Installation

git clone https://github.com/riffatnaz/MINT.git
cd MINT

Then, in R:

install.packages(c("tidyverse", "survival", "ashr", "survminer", "msigdbr", "PoiClaClu", "rmarkdown", "flextable"))
if (!require("BiocManager", quietly = TRUE)) install.packages("BiocManager")
BiocManager::install(c("DESeq2", "TCGAbiolinks", "SummarizedExperiment", "clusterProfiler", "enrichplot", "ReactomePA", "EnhancedVolcano", "org.Hs.eg.db", "apeglm", "fgsea", "pheatmap"))

Usage

1. Open config.R and set your TCGA project ID, significance cutoffs (PVAL_CUTOFF, LFC_CUTOFF), and gene of interest.
2. Run the pipeline:

source("run_pipeline.R")

3. On completion, the HTML report opens automatically from Results/, alongside all figures and result tables generated along the way.

Adapting to a new dataset

MINT was validated on TCGA-BRCA and TCGA-LUSC. To point it at a different TCGA project, change the project ID in config.R — no code changes required. GEO-dataset support is kept in mind structurally but not yet implemented.

Acknowledgements

MINT is built on the shoulders of several open-source R/Bioconductor packages:

• TCGAbiolinks — TCGA data query and download
• DESeq2 — differential expression analysis
• clusterProfiler — GO and KEGG over-representation and GSEA
• ReactomePA — Reactome pathway enrichment
• msigdbr — MSigDB gene sets for GSEA
• fgsea — fast pre-ranked gene set enrichment
• survival / survminer — Kaplan–Meier and Cox proportional hazards modeling
• EnhancedVolcano — volcano plot visualization
• rmarkdown — automated HTML report generation

License

MIT License see LICENSE for details.

Reach Out

Riffat Naz — GitHub — LinkedIn