A DESeq2 tutorial in R, step by step, on real RNA-seq data: which genes change with a steroid, and how to check every step.
We use the airway study (4 donors’ airway muscle cells, untreated or given dexamethasone) and follow one gene, DUSP1, from its raw counts to its line in the results. Six steps: check you have raw counts (not FPKM or TPM); choose the samples for your question; put the donor in the design (~ cell_line + treatment), which takes you from 3,220 genes to 4,743; read the log2 fold change and the adjusted p-value; check with a PCA plot and a volcano plot; and check against the paper, gene by gene, without forcing the numbers to match.
Multilo is our own product. In the last part, one model (GPT-6 Luna) gets the same counts, the sample sheet and the paper, and one prompt. We check its answer against ours, and show its two misses.
✅ Raw counts vs FPKM and TPM
✅ The design line most people get wrong
✅ How to read log2 fold change and adjusted p
✅ PCA and volcano plots as checks
👉 Try Multilo free: 📄 The R scripts, sample sheet and answer key: 🔔 Subscribe for more Multilo tutorials.
⏱️ Chapters
0:00 RNA-seq with DESeq2: which genes changed?
0:31 What DESeq2 gives you: log2 fold change and adjusted p-value
0:54 The example: the airway study (GEO GSE52778)
1:19 Step 1: raw counts, not FPKM or TPM
1:48 Step 2: choose the samples
2:06 Step 3: the DESeq2 design (~ cell_line + treatment)
2:41 Step 4: run DESeq2 and read one line (DUSP1)
3:13 Step 5: PCA plot and volcano plot
3:41 Step 6: check against the paper
4:19 The same data in Multilo (GPT-6 Luna)
5:00 What it got right, and its two misses
5:47 Recap: 6 checks for your RNA-seq analysis
📋 The data (open) Himes BE et al. RNA-Seq transcriptome profiling identifies CRISPLD2 as a glucocorticoid responsive gene… PLOS ONE 2014. Raw counts (NCBI GEO GSE52778):
📋 What I asked Multilo (8 Nov 2026, GPT-6 Luna, one prompt) Using DESeq2 in R, find the genes that change with dexamethasone compared with untreated cells. Use only the untreated and dexamethasone samples, and count a gene as changed if its adjusted p-value is below 0.05. Save the R script, a results table with gene symbols, a PCA plot and a volcano plot, and a short results paragraph in Word that compares the top genes with the paper.
The narration is an AI clone of the presenter’s own voice.
About Multilo Multilo is the AI document IDE (integrated writing environment) for research & academic writing — open Microsoft Word docs, run project-aware AI agents on your real draft, and verify every citation against a real source. Built for students, researchers, and PhDs. 🌐 · 🔔 Subscribe for more Multilo tutorials
#DESeq2 #RNAseq #Bioinformatics #Multilo
source