Bootcamp / Session 4 / 4.3
Session 4.3 · Homework
Does losing liver JNK switch on PPARα targets?
Back to the liver count table from Session 2.4, now with DESeq2. You and an AI agent test which genes change with diet and with each JNK knockout, and check the paper’s main claim yourself.
The data
Vernia et al., 2014, Cell Metabolism (open access) report that liver JNK holds back PPARα, so livers without JNK switch on PPARα target genes such as Fgf21, especially on a high-fat diet. The table is the one from 2.4: RNA-seq expected counts for 24 mouse livers, four genotypes (wt, ALAB, J1c, J2c) on chow or high-fat diet (hfd), three mice each. Sample names are <diet>.<genotype>.rep<n>.
Set up
Make the homework folder and download the table
macOS:
mkdir -p ~/Documents/bootcamp/session4/homework cd ~/Documents/bootcamp/session4/homework curl -fsSL -o genes_expression_expected_count.tsv https://biocore.umassmed.edu/data/session2/genes_expression_expected_count.tsv
Windows PowerShell:
$hw = Join-Path ([Environment]::GetFolderPath('MyDocuments')) 'bootcamp\session4\homework'
New-Item -ItemType Directory -Force $hw | Out-Null
cd $hw
curl.exe -fsSL -o genes_expression_expected_count.tsv https://biocore.umassmed.edu/data/session2/genes_expression_expected_count.tsv
Then in VS Code choose File, Open Folder, and open homework, as you did for RNA-Seq in 4.2. Keep the agent in the same ask-first mode.
The homework
Give the agent one prompt at a time. Read each answer before the next.
1. What does the diet do?
genes_expression_expected_count.tsv in this folder has RNA-seq expected counts: columns gene, transcript, then 24 samples named <diet>.<genotype>.rep<n>. Do not open or print it. Write a Python script that uses pydeseq2, run it with uv, and save results as files here. Sum rows by gene name and round, keep genes with at least 10 counts in some sample. Using only the 6 wt samples, test hfd against chow. Tell me how many genes go up and down at padj < 0.05 and |log2 fold change| > 1, and make a volcano plot.
2. What does each knockout do?
Now fit one DESeq2 model on all 24 samples with one group per diet and genotype. Test ALAB, J1c and J2c against wt, on chow and on hfd separately. Give me a table of up and down genes for these 6 tests, with the same cutoffs.
3. Test the paper’s claim
Plot normalized counts for Fgf21, Cyp4a14, Cyp4a10, Hmgcs2 and Acot1, all 8 groups side by side, with each mouse as a point. For each gene, tell me in which genotypes and on which diet it is higher than in wt, using the tests from step 2.
4. Stretch: does JNK change the diet response?
Fit ~ genotype + diet + genotype:diet with wt and chow as references. How many genes have a significant ALAB:hfd interaction? Plot two of them and explain in plain words what the interaction means for each.
- Which changes more genes in this experiment, diet or genotype?
- Which knockout has the largest effect, and does that fit the gene checks from 2.4 (Mapk8 for JNK1, Mapk9 for JNK2)?
- Do the PPARα targets support the paper’s claim? On chow, on high-fat diet, or both?
- Each group has 3 mice. Which of your conclusions would you trust least?
Links
After the session
Two minutes of feedback, so the next session is pitched better. It also records your attendance, so fill it in whether you were in the room, on Zoom, or catching up on the recording.