DESandbox

Unified Differential Expression Analysis in R

Compare DESeq2, edgeR, and limma-voom results with one simple interface.

✨ Why DESandbox?

• Easy to use: Same commands for all three DE methods
• Compare results: See how different methods agree
• Beautiful plots: Publication-ready visualizations
• Find consensus: Identify genes supported by multiple methods

📦 Installation

# Install dependencies
if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install(c("DESeq2", "edgeR", "limma"))

# Install DESandbox (coming soon on Bioconductor)
devtools::install_github("jleebio/DESandbox")

🚀 Getting Started

Quick start with the airway example (no vignette needed):

# Install required packages if needed
if (!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocManager")
BiocManager::install(c("airway", "DESeq2", "edgeR", "limma"))

library(DESandbox)
library(airway)
library(SummarizedExperiment)

# Load example data
data("airway", package = "airway")

# Prepare counts and metadata
counts <- as.matrix(assay(airway))
metadata <- as.data.frame(colData(airway))
metadata$dex <- stats::relevel(factor(metadata$dex), ref = "untrt")

# Create DESandbox object
dso <- create_desandbox_object(counts, metadata, condition_column = "dex")

# Run DESeq2 (you can also run edgeR/limma-voom)
de_deseq2 <- run_deseq2(dso, alpha = 0.05, lfc_threshold = 1)

# Standardize and plot volcano
std_df <- standardize_results(de_deseq2)
std_df$gene_id <- std_df$gene
std_df$log2FC  <- std_df$log2FoldChange
plot_volcano(std_df, facet_by_method = FALSE, label_top_n = 10,
             title = "airway: DESeq2 Volcano")

📚 Learn More

See the vignettes folder for detailed guides:

• getting_started.Rmd - Complete workflow example
• Additional guides in inst/doc/

🎯 Quick Example (Run All Methods)

Prefer to compare methods in one go?

results <- run_all_methods(dso)            # runs DESeq2, edgeR, limma-voom
comparison <- compare_methods(results)     # overlaps and consensus

# Example: plot DESeq2 volcano from combined results
de2_std <- standardize_results(results$DESeq2)
de2_std$gene_id <- de2_std$gene; de2_std$log2FC <- de2_std$log2FoldChange
plot_volcano(de2_std, facet_by_method = FALSE)

What you can learn from comparison:

# 1) Per-method summary (counts of significant/up/down genes)
comparison$summary_stats

# 2) Consensus genes (significant in all methods)
head(comparison$consensus_genes, 10)

# 3) Method-specific genes (unique to each method)
sapply(comparison$method_specific_genes, length)

# 4) Pairwise agreement (Jaccard index 0–1)
round(comparison$jaccard_matrix, 2)

# 5) Visualize overlaps with a Venn diagram (2–3 methods)
de2 <- standardize_results(results$DESeq2); de2$method <- "DESeq2"
edg <- standardize_results(results$edgeR);  edg$method <- "edgeR"
lim <- standardize_results(results$`limma-voom`); lim$method <- "limma-voom"

# Adapt columns required by plot_venn_de
adapt <- function(df) { df$gene_id <- df$gene; df$log2FC <- df$log2FoldChange; df }
venn_df <- rbind(adapt(de2), adapt(edg), adapt(lim))

p_venn <- plot_venn_de(venn_df, log2FC_threshold = 1, padj_threshold = 0.05,
                       methods = c("DESeq2", "edgeR", "limma-voom"),
                       title = "DE gene overlap (airway)")
print(p_venn)

🔗 Links

• Documentation: See vignettes/ folder (or use the quick start above)
• Report Issues: GitHub Issues
• Bioconductor: Coming soon

📄 License

MIT License - See LICENSE file

---

Have questions? Check the vignettes first - they cover everything!