🧬 TCGA-BRCA RNA-seq Differential Expression and Machine Learning Analysis

 ___ ___ ___    _      ___ ___  _____   _
| _ )_ _/ __|  /_\    | _ )  \ / / __| / |
| _ \| | (__  / _ \   | _ \ |) | (__   | |
|___/___\___|/_/ \_\  |___/___/ \___|  |_|

Ahmed Mohsin Ali¹ ¹Department of Computer Science, Jamia Millia Islamia, Jamia Nagar, New Delhi, India, 110025

"2,384 genes whispering the same story — from raw counts to a 100%-accurate signature."

A full transcriptomic pipeline on TCGA-BRCA: differential expression, functional enrichment, and machine learning–based biomarker discovery in breast invasive carcinoma.

🔬 Overview

213 breast tissue samples. 60,660 genes. One question: what separates tumor from normal at the transcriptome level — and can a handful of genes predict it?

This pipeline runs the full arc from raw RNA-seq counts to a validated, ML-ranked biomarker signature: quality control → normalization → differential expression → functional enrichment → classification.

🎯 The headline result: a 10-gene signature classifies tumor vs. normal with up to 100% accuracy (LASSO) and AUC ≈ 1.0 across all three models tested.

🗺️ The Pipeline at a Glance

flowchart LR
    A["📥 Data Acquisition<br/>TCGA-BRCA, STAR counts"] --> B["🧹 Quality Control<br/>STAR two-pass, GRCh38"]
    B --> C["⚖️ Normalization<br/>DESeq2 + TPM"]
    C --> D["📉 Differential Expression<br/>Tumor vs Normal"]
    D --> E["🧭 Functional Enrichment<br/>GO / KEGG / DO"]
    D --> F["🤖 Machine Learning<br/>RF, LASSO, SVM"]
    E --> G["📈 Biomarker Signature"]
    F --> G

🧫 Dataset

Detail	Value
Source	TCGA-BRCA (RNA-seq, STAR counts)
Tumor / Normal / Total samples	100 / 113 / 213
Genes (raw)	60,660
Genes tested (post-filter)	25,571

📊 Key Results

By the Numbers

🧬 Significant DEGs	⬆️ Upregulated	⬇️ Downregulated	🎯 Best Accuracy	📐 Best AUC
2,384	1,135	1,249	100% (LASSO)	≈1.0

Top Enriched Pathways

Cell cycle regulation · DNA replication · PI3K–Akt signaling · p53 signaling — all hallmarks of breast tumor progression.

Model Showdown

Model	Accuracy	Sensitivity	Specificity	AUC
Random Forest	0.9683	0.9394	1.000	1.000
🏆 LASSO	1.0000	1.0000	1.000	1.000
SVM	0.9841	0.9697	1.000	1.000

Top Biomarker Candidates

VEGFD · MMP11 · COL10A1 · UBE2T · NEK2 — ranked by combined Random Forest importance and LASSO coefficient.

📁 Repository Structure

tcga-brca-rnaseq-degs-ml/
├── scripts/
│   └── tcga_brca_rnaseq_analysis.R
├── figures/
├── tables/
├── sessionInfo.txt
├── LICENSE
└── README.md

⚙️ Requirements

Tool	Role
DESeq2	Normalization & differential expression
clusterProfiler	GO / KEGG / Disease Ontology enrichment
caret, randomForest, glmnet	Machine learning (RF, LASSO)
pROC	ROC / AUC evaluation
ggplot2, pheatmap, ComplexHeatmap	Visualization

install.packages("BiocManager")
BiocManager::install(c("DESeq2","edgeR","limma","clusterProfiler","enrichplot",
                       "DOSE","org.Hs.eg.db","TCGAbiolinks","ComplexHeatmap",
                       "pathview","biomaRt"))
install.packages(c("ggplot2","pheatmap","caret","randomForest","glmnet",
                    "pROC","tidyverse"))

🚀 How to Run

git clone https://github.com/amuhsenali/tcga-brca-rnaseq-degs-ml.git
cd tcga-brca-rnaseq-degs-ml
Rscript scripts/tcga_brca_rnaseq_analysis.R

📦 Raw TCGA-BRCA data isn't redistributed here — the script pulls it directly via TCGAbiolinks from the GDC Data Portal.

📖 Citation

Ali, A. M. (2026). TCGA-BRCA RNA-seq Differential Expression and Machine Learning Analysis [Computer software]. GitHub. https://github.com/amuhsenali/tcga-brca-rnaseq-degs-ml

📄 License

MIT License — see LICENSE.

✉️ Contact

Ahmed Mohsin Ali

FilesExpand file tree

README.md

Latest commit

History