scvi-tools Single-Cell Modeling Skill
A reference skill that guides Claude through scvi-tools deep generative models for batch correction, multimodal integration, and probabilistic differential expression in single-cell omics.
Data & AnalyticsAdvanced★ 33,030⑂ 3,248AI score 7/10Last updated: Aug 9, 2026
What it does
scvi-tools is a PyTorch/Lightning framework of probabilistic models for single-cell genomics. This skill gives Claude a curated map of the library:
- Model selection by modality: scRNA-seq (scVI, scANVI, AUTOZI, VeloVI, contrastiveVI), ATAC-seq (PeakVI, PoissonVI, scBasset), multimodal (totalVI, totalANVI, MultiVI, MrVI, DIAGVI), spatial (DestVI, Stereoscope, Tangram, scVIVA), specialized (MethylVI, CytoVI, Solo, CellAssign)
- Namespace clarity: core models under
scvi.model, everything else underscvi.external— prevents a very common import mistake - Canonical workflow:
setup_anndata→train→get_latent_representation/get_normalized_expression→ scanpy neighbors/UMAP/Leiden - Uncertainty-aware DE:
differential_expression(mode="change", delta=...) - Best practices: raw counts only, gene filtering, covariate registration, model persistence, GPU acceleration
- Pointers to deeper
references/files per model family and to theoretical foundations
Who it's for
- Computational biologists and grad students working with single-cell RNA/ATAC data
- Teams integrating many batches or studies where basic scanpy correction falls short
- Anyone modeling CITE-seq, multiome, or spatial data probabilistically
- Overkill if you only need a standard scanpy QC-to-clustering pipeline
Example uses
- "Batch-correct this multi-donor scRNA-seq object with scVI and give me a UMAP" → produces filtering,
setup_anndatawith batch/categorical/continuous covariates, training,X_scVI, and downstream scanpy calls - "I have CITE-seq protein + RNA — which model?" → recommends totalVI (or totalANVI when labels exist) with working code
- "Run differential expression between two cell types with uncertainty" → uses
mode="change"with adeltaeffect-size threshold and explains how to read the output
· · · Install guide · · ·
Install in the Claude app (no terminal)
- Download the ZIP with the button below.
- In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
- Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
Install in Claude Code
Let Claude do it — paste this into Claude Code
Install the skill I found on Claude Skill Mart. Copy the skills/scvi-tools folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/scvi-tools/. When it's done, tell me in one line what this skill can do.
Install with a command instead
git clone https://github.com/K-Dense-AI/scientific-agent-skills.git && mkdir -p ~/.claude/skills && cp -r scientific-agent-skills/skills/scvi-tools ~/.claude/skills/⚠ This is a third-party skill. Check the source repository before installing.
- Open a terminal and clone the repository:
git clone https://github.com/K-Dense-AI/scientific-agent-skills.git - Create your skills folder if needed:
mkdir -p ~/.claude/skills - Copy just this skill:
cp -r scientific-agent-skills/skills/scvi-tools ~/.claude/skills/ - Confirm that
SKILL.mdand thereferences/directory both landed in~/.claude/skills/scvi-tools/. - Install the library in a Python 3.12+ environment:
uv pip install scvi-tools(oruv pip install "scvi-tools[cuda]"for GPU). - Restart Claude Code and prompt something like "use scvi-tools to batch-correct this AnnData object".
- The doc cites a version/date that may be inaccurate, so verify with
python -c "import scvi; print(scvi.__version__)"and cross-check docs.scvi-tools.org before relying on any specific API.
View source on GitHub ↗License: MIT