bulk-rnaseq-counts-to-de-deseq2

DESeq2 Comprehensive Reference

Complete code patterns for DESeq2 differential expression analysis. Adapt these examples to your experimental design.

Decision-making: see decision-guide.md | Errors: see troubleshooting.md

Complete Standard Workflow

library(DESeq2)
library(apeglm)

# 1. Create DESeqDataSet
dds <- DESeqDataSetFromMatrix(countData = counts, colData = coldata, design = ~ condition)

# 2. Pre-filter low counts
keep <- rowSums(counts(dds)) >= 10
dds <- dds[keep,]

# 3. Set reference level
dds$condition <- relevel(dds$condition, ref = 'control')

# 4. Run DESeq2 pipeline
dds <- DESeq(dds)

# 5. Extract results
res <- results(dds)

# 6. Apply LFC shrinkage
resLFC <- lfcShrink(dds, coef = 'condition_treated_vs_control', type = 'apeglm')

# 7. Get significant genes
sig <- subset(res, padj < 0.05 & abs(log2FoldChange) > 1)

Design Formulas

Simple Two-Group

design = ~ condition

Use: Single factor, no batch effects, most common starting point.

Batch Correction

design = ~ batch + condition

Use: Multiple sequencing runs, PCA shows batch clustering. Requirement: each condition must have samples in each batch (not confounded).

Paired Samples

design = ~ individual + condition

Use: Before/after treatment, tumor vs normal from same patient. Benefit: controls individual variation, increases power.

Interaction

design = ~ genotype * treatment
# Expands to: ~ genotype + treatment + genotype:treatment

Use: Test if treatment effect differs by genotype/sex/age.

Extract results:

res_interaction <- results(dds, name = "genotypeMutant.treatmentdrug")
res_treatment_WT <- results(dds, name = "treatment_drug_vs_control")

Multi-Factor

design = ~ sex + age_group + treatment

Use: Multiple confounders to adjust for. Requirement: ≥3 samples per coefficient, variables not confounded.

No-Intercept

design = ~ 0 + group

Use: Direct comparisons between any groups.