This repository is a Cursor skill and a small Python plotting library. Use it in a new bioinformatics project when you have tidy tables from several experiments, datasets, or models and you need a figure that matches the scientific question.
It recommends a plot family and can draw the renderers listed in docs/DATA_SCHEMAS.md. It does not download papers or run DESeq2, a Cox model, or a mixed model. Atlas template grades are visual preferences, not proof that a test fits your study.
Start here: docs/QUICK_START.md.
- Full guide: docs/USAGE_GUIDE.md
- Columns and renderer mismatches: docs/DATA_SCHEMAS.md
- Copy-ready prompts: docs/PROMPT_LIBRARY.md
- Skill instructions: .cursor/skills/bioinformatics-visualization/SKILL.md
git clone https://github.com/CyrilleMesue/bioinformatics-visualization-skill.git
cd bioinformatics-visualization-skill
python -m pip install -r requirements.txt
export PYTHONPATH=src
python -c "from bioinformatics_visual_evidence_atlas.visualization import recommend; print('ok')"Copy .cursor/skills/bioinformatics-visualization into the project where you will chat. There is no [build-system] entry, so do not rely on pip install -e ..
Synthetic rows in data_templates/group_comparison_long.csv. One row is one sample.
Primary question: Does the treated group differ from control?
File: data_templates/group_comparison_long.csv
One row is one independent sample. Columns: group, value.
Panel mode: single.
Panel modes (single, auto, specified) and the three candidate designs are instructions to the agent. They are not function arguments. figures.render writes PNG, SVG, PDF, plotting_data.csv, caption.txt, and audit.json.
Implemented: routing, template lookup from milestone_8/template_rankings.json, and the renderers in visualization/figures.py. Agent-guided: panel planning, candidate comparison, and an accessibility write-up. Not implemented: paired plots, UMAP, Cox models, and a volcano that reads DESeq2 output. Details are in the usage guide.
Do not fabricate microscopy, blots, or molecular structures. Do not present a synthetic figure as a published result.
These are synthetic reproductions graded 10. Standalone panels and grouped panels are separate. The full set, every grade 9 and 10 plus a grade 8 when that is the top grade in its family, is in .cursor/skills/bioinformatics-visualization/examples/EXAMPLES.md.
After a figure is drawn, compare it with the examples for that family. The result should be as clear: readable type, panel letters free of titles and legends, visible uncertainty, and space between panels. A grade 10 is not automatic. Use a grade 9 example when its layout stays readable and the grade 10 layout is crowded.
