DESeq2 vs edgeR Benchmark on Brown Bear RNA-seq (Hibernation Study)

This repository contains my complete analysis comparing DESeq2 and edgeR on the same RNA-seq dataset to answer a commonly asked question:

“Why choose DESeq2 over edgeR (or vice-versa)?”

Instead of relying on theory, I tested both tools head-to-head using real RNA-seq data.

🧬 Dataset

• Accession E-MTAB-14479

• Organism: Brown Bear (Ursus arctos)

• Condition: Hibernation vs Active (pre-hibernation)

• Source: ArrayExpress raw count matrix

• Type: Bulk RNA-seq

• Design: Two-group comparison (Hibernation vs Awake)

📦 Tools Compared

• DESeq2

• edgeR

Both were run independently from raw counts → normalization → dispersion estimation → DE analysis.

📊 Key Results

Metric	DESeq2	edgeR
Total DEGs (FDR < 0.05)	8,964	8,324
Upregulated	4,463	4,113
Downregulated	4,501	4,216
Common DEGs	8,275	—

Correlations Between Methods

• log2FC correlation: 0.9997

• Adjusted p-value correlation: 0.9973

• Top-100 DEG overlap: 80 / 100

• Sign concordance: 50.86%

• rlog vs logCPM correlation: 0.9901

• Dispersion trend correlation: 0.93

Both methods showed extremely high agreement on the biological signal.

🧠 When Does Each Tool Shine?

✔️ Choose DESeq2 if your dataset has:

• Moderate sample size

• Higher biological variation

• Need for strong fold-change shrinkage

• Need for slightly higher sensitivity

• Preference for rlog/VST stabilization

✔️ Choose edgeR if your dataset has:

• Very small sample size (n = 2–3 per group)

• Many highly expressed genes

• Need for conservative/strict DEG lists

• Preference for speed and efficiency

• Desire for a cleaner, tighter DEG set

---

📫 Contact

📧 Feel free to connect or reach out:
🔗 🌐 LinkedIn | 📫 gunjansarode.bioinfo@gmail.com
🐛 Found a bug or have a question? Open an issue on this repo!

---