🧬 Compare Gene Expression Differential Expression Analysis in R

Eastern European Bioinformatics and Computational Genomics Workshop (EEBG 2026)

EEBG 2026 - Differential Expression Analysis (DEA) in R

Author: Tomek Gaczorek (Jagiellonian University)

Project Overview

This repository contains a comprehensive workflow for performing Differential Expression Analysis (DEA) on bulk RNA-seq data using R. The primary objective of this analysis is to identify differentially expressed genes between smoker and non-smoker groups, analyzing male and female cohorts separately based on the GSE237252 dataset.

Dataset

The analysis utilizes the following data files:

• GSE237252_raw_data.csv: Raw gene expression counts.
• GSE237252_males_metadata.csv: Metadata for the male cohort, detailing smoking status.
• GSE237252_females_metadata.csv: Metadata for the female cohort, detailing smoking status.

Dependencies

The analysis is conducted in R (v4.6.1) and relies on the following key Bioconductor and CRAN packages:

• Data manipulation: dplyr
• DEA core: DESeq2
• Visualization: ggplot2, enrichplot
• Functional Analysis: clusterProfiler, org.Hs.eg.db

---

Workflow Summary

1. Data Preprocessing

• Raw counts are loaded and matched to corresponding metadata.
• Lowly expressed genes are filtered out based on a threshold (e.g., rowSums > 42).

2. Differential Expression Analysis (DESeq2)

• Model Design: ~SMOKING (comparing smokers vs. nonsmokers).
• Standard DESeq2 pipeline steps: estimating size factors, estimating gene-wise dispersions, and fitting the negative binomial GLM.
• Log2 fold-change (LFC) shrinkage is applied to handle noise in low-count genes.

3. Quality Control & Exploratory Data Analysis

Several visual diagnostic tools are implemented to inspect the integrity of the data and normalization:

• Raw vs. Normalized Counts: Evaluates the normalization efficacy.

• Dispersion Estimates: Checks the mean-dispersion relationship.

• Principal Component Analysis (PCA): Visualizes sample clustering by variance (PC1 & PC2) to observe group separations.

4. Result Visualization

Significant genes are identified (using thresholds like padj < 0.01 and |log2FoldChange| > 2) and visualized using:

• Log2-fold Shrinkage Density Plots: Comparing original vs. shrinked distributions.

• MA Plot: Showing log-fold change against the mean of normalized counts.

• Volcano Plot: Highlighting highly significant and highly folded-changed genes.

5. Functional Enrichment Analysis

• Over-representation analysis of Gene Ontology (GO) Biological Processes (BP) is performed using clusterProfiler on the significantly differentially expressed genes.
• The top enriched pathways are visualized using dot plots.

(Note: Add the GO-terms dotplot to the male_results folder and link it here when generated)

---

Female Cohort Analysis

The repository also includes a complete replication of the pipeline specifically on the female cohort. This allows for an independent assessment and a cross-sex comparison of the transcriptomic impacts of smoking.

(Note: Ensure female plots are saved in the female_results directory using the same naming conventions to render correctly)

Usage

To reproduce the analysis:

1. Clone the repository.
2. Ensure the GSE237252 raw data and metadata CSV files are located in your working directory.
3. Open the main R script / Jupyter Notebook.
4. Run the chunks sequentially to generate the DESeq2 objects, plots, and enrichment tables.