RNA-seq DESeq2 Nextflow Pipeline on AWS

End-to-end automated RNA-seq differential expression pipeline built with Nextflow DSL2, running on AWS EC2/Batch with S3 for data storage.

Overview

This pipeline performs differential expression analysis of RNA-seq count data using DESeq2, with full automation from data download through results upload to S3.

Biological application: Alzheimer's disease mouse model transcriptomics
Dataset: 3xTgAD vs WT cortex (GEO: GSE161904)
Author: Ravi Shanker Pandey, Ph.D. | The Jackson Laboratory

Pipeline Architecture

S3 (input counts matrix) ↓ DOWNLOAD_COUNTS — pulls data from AWS S3 ↓ RUN_DESEQ2 — differential expression analysis ↓ UPLOAD_RESULTS — pushes results back to S3

Key Features

• Nextflow DSL2 — modular, scalable pipeline architecture
• AWS integration — EC2 compute, S3 storage, Batch-ready
• Automated DESeq2 — complete differential expression workflow
• Gene annotation — ENSEMBL to gene symbol mapping via org.Mm.eg.db
• Publication-quality plots — volcano plots with AD-relevant gene labels
• Reproducible — Docker/Singularity compatible, timestamped logging

Pipeline Components

File	Description
main.nf	Main Nextflow DSL2 workflow
nextflow.config	Pipeline configuration (local + AWS Batch profiles)
bin/deseq2_analysis.R	DESeq2 differential expression R script

Requirements

• Nextflow >= 23.0
• Java >= 17
• R >= 4.3 with DESeq2, ggplot2, org.Mm.eg.db
• AWS CLI configured with S3 access
• Docker (for containerized execution)

Usage

Local execution

nextflow run main.nf -profile local

AWS Batch execution

nextflow run main.nf -profile aws

Custom parameters

nextflow run main.nf -profile local \
  --bucket your-s3-bucket \
  --counts_file path/to/counts.txt \
  --pval_cutoff 0.05 \
  --lfc_cutoff 0.5

Results

The pipeline generates:

• all_genes.csv — DESeq2 results for all tested genes
• significant_genes.csv — Filtered significant genes (padj < 0.05)
• volcano_plot.png — Publication-quality volcano plot

Biological Results

Applied to 3xTgAD vs WT cortex data (30 samples, 43,629 genes):

• 2,339 significant genes identified (padj < 0.05)
• Upregulated in AD: Spp1, Lpl, Tyrobp, Hexb (neuroinflammation)
• Downregulated in AD: Bin1, Camk2a (synaptic dysfunction)

Author

Ravi Shanker Pandey, Ph.D.
Associate Computational Scientist, The Jackson Laboratory
ORCID | GitHub | LinkedIn