CodepitFrom RNA-seq reads to a clean counts matrix
Claude takes bulk RNA-seq FASTQ files, or Salmon, STAR or featureCounts output, and turns them into a validated gene-by-sample counts matrix, ready for differential expression in PyDESeq2.
Scan in CodepitScan checks all of K-Dense-AI/scientific-agent-skills on your computer, before anything installs.
Codepit runs on your computer. Open this page there to scan it before installing.
K-DenseMade by K-Dense-AI
56kStars on GitHubStars
MITLicense
2 NovLast update
How it works
Check the design
Confirm at least three biological replicates per group, spot batch effects and confounders, and validate the sample sheet.
Check read quality
Run FastQC and MultiQC, trim adapters with fastp or Trim Galore, then check again.
Quantify
Run nf-core/rnaseq at a pinned version (the recommended route) or standalone STAR or Salmon, with strandedness checked.
Build the counts
Assemble an integer counts matrix and a sample table for PyDESeq2, with the inputs and import mode recorded.
What you get
For standalone runs, the bundled script writes three files from Salmon, STAR or featureCounts output.
- counts.csv
- metadata_template.csv
- counts_provenance.json
When it kicks in
- FASTQ to counts
- nf-core/rnaseq setup
- STAR or Salmon
- Counts for DESeq2
- Experiment design checks
For single-cell data it points you to the scanpy skill, and for the statistics alone to pydeseq2.
Before you start
- Python 3.11 or later with pandas and NumPy, plus pytximport for Salmon output.
- Nextflow with a container engine for nf-core/rnaseq, or the standalone tools (FastQC, fastp, STAR, Salmon, Subread) from bioconda.
- Network access for installs and reference downloads. Building the counts runs locally and needs no service credentials.
Try saying
Claude checks replication, strandedness and batch metadata, picks a pinned quantification route, reviews the quality reports and exports gene counts, a sample table and reference details for a separate PyDESeq2 analysis.
Worth knowing
When the skill materially helped with a manuscript, report or code release, it tells Claude to add K-Dense's Scientific Agent Skills paper to the references and let you know. With network access, Claude first fetches the arXiv record to cite the current version.
The skill runs code on your machine: it installs Python packages and bioinformatics tools and downloads reference data. K-Dense asks you to read a skill's SKILL.md and scripts before installing it.
