miRNA Differential Expression Analysis

A modular and reproducible R workflow for exploratory analysis and differential expression of miRNA count data.

The pipeline supports:

• Quality control and exploratory analysis
• Count filtering and TMM normalization
• Principal Component Analysis (PCA)
• Hierarchical clustering
• Expression distribution assessment
• Library size evaluation
• miRNA abundance visualization
• Differential expression analysis using the limma-voom framework
• Multiple experimental designs through configuration files
• Reproducible outputs and intermediate objects

---

Overview

This repository provides a configurable workflow for the analysis of miRNA count matrices obtained from sequencing or public repositories such as GEO.

The analysis is divided into two independent stages:

1. Exploratory analysis
2. Differential expression analysis

Configuration is handled through a single config.yml file, allowing the same codebase to be reused across different datasets and experimental designs.

---

Repository Structure

.
├── config.yml
│
├── data/
│   └── example/
│       ├── example_counts.csv
│       └── example_metadata.csv
│
├── scripts/
│   ├── 00_download_geo_data.R
│   ├── exploratory_analysis.R
│   └── differential_expression.R
│
├── R/
│   ├── load_raw_data.R
│   ├── dge.R
│   ├── pca.R
│   ├── clustering.R
│   ├── distributions.R
│   ├── library_size.R
│   ├── abundance.R
│   ├── diff_expression.R
│   └── utils.R
│
├── results/
│
└── README.md

---

Input Data

The workflow requires two input files.

Counts Matrix

A CSV file containing raw miRNA counts.

Example:

gene_id	sample1	sample2	sample3
hsa-miR-21	125	98	143
hsa-miR-16	455	390	502

Rows correspond to miRNAs and columns correspond to samples.

---

Metadata

A CSV file describing samples and experimental variables.

Example:

id	trimester	fetal_sex
GSM1	First	Female
GSM2	Third	Male

The id column must match the sample names present in the count matrix.

---

Configuration

All analysis parameters are specified in config.yml.

Example:

metadata_file: data/example/example_metadata.csv
counts_matrix_file: data/example/example_counts.csv

all_variables:
  - trimester
  - fetal_sex

interest_variable: trimester

interest_pcs:
  - PC1
  - PC2

distances:
  - corr
  - euclidean

clustering_methods:
  - complete
  - average

formulas:
  trimester: "~ 0 + trimester"
  sex_trimester: "~ 0 + trimester + fetal_sex"

This design allows exploratory and differential expression analyses to be performed without modifying the source code.

---

Exploratory Analysis

The exploratory workflow performs the following analyses.

Filtering

Lowly expressed miRNAs are removed using:

filterByExpr()

---

Normalization

Library size normalization is performed using the TMM method:

calcNormFactors()

Log2 CPM values are subsequently computed for downstream analyses.

---

Library Size Assessment

Raw and filtered library sizes are evaluated for each sample.

Outputs:

results/exploratory_analysis/library_sizes/

---

Expression Distributions

Per-sample expression distributions are visualized using normalized log2 CPM values.

Outputs:

results/exploratory_analysis/distributions/

---

Principal Component Analysis (PCA)

PCA is performed on normalized expression values.

Outputs:

results/exploratory_analysis/pca/

---

Hierarchical Clustering

Hierarchical clustering can be performed using:

• Pearson correlation distance
• Euclidean distance

Supported linkage methods:

• complete
• average

Outputs:

results/exploratory_analysis/clustering/

---

miRNA Abundance

Visualization of the most abundant miRNAs across samples and experimental groups.

Outputs:

results/exploratory_analysis/abundance/

---

Differential Expression Analysis

Differential expression analysis is performed using the limma-voom framework.

Workflow:

voom()
lmFit()
contrasts.fit()
eBayes()

The design matrix is automatically generated from formulas defined in config.yml.

Multiple models can therefore be tested without modifying the analysis code.

Example:

formulas:
  trimester: "~ 0 + trimester"
  sex_trimester: "~ 0 + trimester + fetal_sex"

---

Differential Expression Outputs

For each contrast, the workflow generates:

Complete Differential Expression Table

DE_*.csv

Contains all tested miRNAs.

---

Significant Differentially Expressed miRNAs

DEG_*.csv

Contains significant miRNAs after multiple testing correction.

Significance criteria:

• Adjusted p-value < 0.05
• |logFC| > 1

---

Volcano Plots

DEG_*_volcano.png

Visualization of differential expression results.

---

Heatmaps

DEG_*_heatmap.png

Expression heatmaps of significant miRNAs.

---

Running the Workflow

Exploratory Analysis

Rscript scripts/exploratory_analysis.R

---

Differential Expression Analysis

Rscript scripts/differential_expression.R

---

GEO Example Dataset

An example GEO dataset can be downloaded using:

Rscript scripts/00_download_geo_data.R

The script generates:

data/example/
├── example_counts.csv
└── example_metadata.csv

which can be used directly to reproduce the analyses contained in this repository.

---

Intermediate Objects

The workflow stores reusable intermediate objects:

results/rds/
├── dge.rds
└── log2_cpm.rds

These files allow downstream analyses to be performed without repeating filtering and normalization steps.

---

Main Dependencies

CRAN:

• tidyverse
• ggplot2
• pheatmap
• ggdendro
• cowplot
• ggpubr
• yaml

Bioconductor:

• edgeR
• limma
• GEOquery

---

Citation

If you use this repository, please cite:

Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics (2010).

Ritchie ME et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research (2015).

---

License

This project is distributed under the terms of the MIT License.