Bulk_RNA_Seq_Analysis

Reproducible repository for bulk RNA-seq analyses (colon organoids and keratinocyte cell lines) – from my undergrad 2nd-3rd year (2022-2023)

Project summary

This repository contains scripts, metadata, raw and processed count tables, and results for differential expression analyses and downstream visualizations. The recent reorganization added a recommended folder layout, helper files for reproducible environments, and a .gitignore tuned to ignore intermediate data files.

Repository Structure

.
├── colon organoids/      # Analysis scripts and data for colon organoid experiments
├── keratinocyte/         # Analysis scripts and data for keratinocyte cell lines
├── pipeline/             # Reusable pipeline scripts for RNA-seq analysis
├── data/                 # Raw and processed data files
├── results/              # Output files from analyses
├── scripts/             # Utility scripts and shared functions
├── requirements.txt     # Python package dependencies
└── R_requirements.R    # R package dependencies

From SRA / GEO FASTQ to gene-level counts (example)

The following shell snippet shows a common end-to-end workflow: fetch FASTQ from SRA/GEO, optionally trim, align to a reference, and generate gene-level counts with featureCounts. There is an alternative Salmon (quasi-mapping) example included.

Prerequisites

• Bash tools for FASTQ to counts:
  • SRA Toolkit (prefetch, fasterq-dump): download FASTQ from SRA/GEO
  • FASTQC: quality control of raw FASTQ files
  • fastp: trimming and filtering reads
  • HISAT2 or STAR: alignment to reference genome
  • samtools: BAM file manipulation
  • subread (featureCounts): gene-level counting
  • Salmon: fast transcript quantification (alignment-free)

Install commands (Ubuntu/Debian):

sudo apt-get update
sudo apt-get install fastqc hisat2 star samtools
# SRA Toolkit
conda install -c bioconda sra-tools
# fastp
conda install -c bioconda fastp
# subread (featureCounts)
conda install -c bioconda subread
# Salmon
conda install -c bioconda salmon

Or use mamba for faster conda installs.

Role in pipeline:

• FASTQC: run after download/trimming for QC reports
• fastp: trim/filter before alignment
• HISAT2/STAR: align reads to genome
• samtools: sort/index BAM files
• featureCounts: count reads per gene
• Salmon: alternative quantification (transcript/gene)

Example: end-to-end (HISAT2 + featureCounts)

# variables
SRR=SRR22450503   # replace with SRA run accession 
THREADS=8
REF_FA=/path/to/genome.fa
GTF=/path/to/annotations.gtf
HISAT2_INDEX=/path/to/hisat2_index_prefix
OUTDIR=data/raw/${SRR}
mkdir -p ${OUTDIR}

# 1) fetch FASTQ
prefetch ${SRR}
fasterq-dump --split-files --outdir ${OUTDIR} ${SRR}

# 2) trim (optional)
fastp -i ${OUTDIR}/${SRR}_1.fastq -I ${OUTDIR}/${SRR}_2.fastq \
			-o ${OUTDIR}/${SRR}_1.trim.fastq -O ${OUTDIR}/${SRR}_2.trim.fastq \
			-w ${THREADS} -h ${OUTDIR}/${SRR}.fastp.html -j ${OUTDIR}/${SRR}.fastp.json

# 3) align with HISAT2
hisat2 -p ${THREADS} -x ${HISAT2_INDEX} \
	-1 ${OUTDIR}/${SRR}_1.trim.fastq -2 ${OUTDIR}/${SRR}_2.trim.fastq \
	| samtools sort -@ ${THREADS} -o ${OUTDIR}/${SRR}.sorted.bam
samtools index ${OUTDIR}/${SRR}.sorted.bam

# 4) count reads at gene level
featureCounts -T ${THREADS} -a ${GTF} -o ${OUTDIR}/${SRR}.featureCounts.txt ${OUTDIR}/${SRR}.sorted.bam

# The counts file contains gene-level raw counts suitable for DESeq2/edgeR downstream.

Quick alternative: Salmon (fast, alignment-free)

# create salmon index (once)
salmon index -t ${REF_FA} -i salmon_index --type quasi -k 31

# quantify
salmon quant -i salmon_index -l A \
	-1 ${OUTDIR}/${SRR}_1.trim.fastq -2 ${OUTDIR}/${SRR}_2.trim.fastq \
	-p ${THREADS} -o ${OUTDIR}/salmon_${SRR}

# export gene-level counts (tximport in R or use salmon2gene tools)

Notes

• Replace variables (paths, indexes) with your environment values.
• For GEO datasets that provide raw FASTQ links, use wget or curl instead of prefetch.
• If your reads are single-end adjust fasterq-dump and alignment commands accordingly.
• Save these pipeline scripts in pipeline/ (created below) and run from project root so outputs go to data/raw/ or data/processed/.

Contents

• colon organoids/, keratinocyte/ — existing analysis folders with R scripts, count matrices and results.
• data/ — recommended place for data; contains raw/ and processed/ subfolders.
• scripts/ — recommended place for analysis and utility scripts.
• results/ — figures, tables and exported outputs.
• R_requirements.R — R and Bioconductor package installer to reproduce the R environment.
• requirements.txt — Python dependencies for any auxiliary scripts.

Getting started

Prerequisites

• R (>= 4.0) and Rtools on Windows if compiling packages.
• Optional: Python 3.8+ if you use Python scripts.

Quick start

1. Clone the repository and change into it:

git clone <repo-url>
cd "RNA Seq Analysis"

2. Install R packages (run inside an R session):

source("R_requirements.R")

3. (Optional) Create and activate a Python virtual environment and install Python dependencies:

python -m venv .venv; .\.venv\Scripts\Activate.ps1; pip install -r requirements.txt

4. Place raw immutable inputs (FASTQ, original count matrices) in data/raw/. Place derived outputs in data/processed/.

Recommended repository layout

• data/raw/ — immutable original inputs.
• data/processed/ — derived tables (normalized counts, filtered results).
• scripts/ — modular, documented scripts; place helper functions in scripts/utils/ as needed.
• results/ — final figures, tables and CSVs used for reporting.

Notes on .gitignore and tracking CSVs

The repository .gitignore was updated to ignore *.csv and common R artifacts to avoid committing large or intermediate tables. Important: files already tracked by git will remain tracked. To stop tracking a file already in the repo, run:

git rm --cached path\\to\\file.csv
git commit -m "Stop tracking large CSV"

If you want specific CSVs to remain tracked, I can add explicit negation rules (e.g., !colon organoids/meta_colon.csv) to the .gitignore.

How to run analyses

• Inspect the R scripts in colon organoids/ and keratinocyte/ to identify pipeline steps.
• Prefer running analyses from the repository root so that relative paths (e.g., data/processed/) resolve correctly.
• Example: run an R script from PowerShell:

Rscript scripts/my_analysis.R

Reproducibility recommendations

• Record package versions using sessionInfo() in R and include this output in result folders.
• Use data/raw vs data/processed/ split to prevent accidental modification of raw inputs.
• Use branches and PRs for large reorganizations.

Contributing

• Open an issue to propose large structural changes.
• Use branches for work and submit PRs with small, reviewable changes.

License & contact

This project uses the license in LICENSE. For questions, open an issue or contact me at joy21.dev.pd@gmail.com.

---