Setup

Five minutes of setup and you’re ready for the whole workshop. If you fall behind on any install, you can still follow along — every result is shown on this site.

1. Install Claude Code

Claude Code is Anthropic’s agentic coding tool that runs in your terminal.

# macOS / Linux
curl -fsSL https://claude.com/install.sh | bash

Then start it and sign in:

claude

See the official docs for other install methods (npm, IDE extensions, the desktop app).

2. Get the workshop repo

git clone https://github.com/crazyhottommy/claude_code_RNAseq_workshop.git
cd claude_code_RNAseq_workshop

The repo already contains the salmon quant.sf files for all four samples in data/salmon/, so you can start analyzing immediately — no large downloads.

data/
├── salmon/
│   ├── Normoxia_sgCTRL_1/quant.sf
│   ├── Normoxia_sgCTRL_2/quant.sf
│   ├── Hypoxia_sgCTRL_1/quant.sf
│   └── Hypoxia_sgCTRL_2/quant.sf
├── samples.csv          # which sample is which condition
├── tx2gene.csv          # transcript → gene
└── gene_name_map.csv    # gene id → gene symbol

3. Install the R / Bioconductor packages

The downstream analysis is in R. Install everything with the helper script:

Rscript install.R

This installs tximport, DESeq2, clusterProfiler, org.Hs.eg.db and the tidyverse pieces (dplyr, readr, ggplot2, pheatmap). Don’t have R? Get it from r-project.org (we use R ≥ 4.3).

4. (Optional) tools for preprocessing

Only needed if you want to regenerate quant.sf from raw FASTQ (covered in Preprocessing). Not required for the live session.

conda env create -f environment.yml   # installs salmon, sra-tools, fastq-dl
conda activate rnaseq

A laptop is plenty: 8 GB RAM minimum (16 GB comfortable), 4+ cores, ~5–30 GB free disk. The compute (index + quant) runs in under ~6 minutes for the read subset; see the full hardware & timing table.

5. Launch Claude Code in the repo

claude

Because the repo ships a CLAUDE.md, Claude Code already knows the dataset, the file layout, and the analysis goal. Try opening with:

Try this prompt

Read CLAUDE.md and the files in data/, then tell me what this dataset is and what analysis we’re about to do.

When it gives a sensible summary, you’re set. → On to 1 · The data.

Back to top