scvi-tools
Guides Claude through scvi-tools deep-learning workflows on AnnData single-cell inputs, writing trained model artifacts, latent representations, and normalized expression matrices.
| Type | Claude Skill |
| Supplier | Anthropic |
| Availability | GA — distributed via anthropics/life-sciences (Oct 2025) |
| Pricing | Free / OSS |
| Capabilities | Read/Write |
How to install
- Also packaged in the SciAgent-Skills collection (jaechang-hits (community OSS, CC BY 4.0)): clone
jaechang-hits/SciAgent-Skillsand run/plugin install sciagent-skillsin Claude Code (or copyskills/genomics-bioinformatics/single-cell/scvi-tools-single-cellinto~/.claude/skills/). - Claude Code — plugin marketplace:
/plugin marketplace add anthropics/life-sciences /plugin install scvi-tools@life-sciences - Claude.ai — Settings → Capabilities → Skills → Upload skill, using the skill bundle from the
anthropics/life-sciencesrepo. - Alternative — K-Dense skill:
npx skills add K-Dense-AI/scientific-agent-skills, then enable thescvi-toolsskill. A community-maintained skill (MIT collection) wrapping the same scvi-tools library; use it if you don’t haveanthropics/life-sciencesmarketplace access.
What it does
Workflows for scVI / scANVI (batch correction, semi-supervised cell-type annotation), totalVI and MultiVI (CITE-seq, RNA+ATAC multi-modal integration), PeakVI and scBasset (chromatin accessibility), DestVI / Tangram / cell2location / Stereoscope (spatial deconvolution), contrastiveVI (perturbation), sysVI (cross-cohort batch correction), and veloVI (RNA velocity). Covers model setup, training, and latent-extraction code patterns.
Primary use cases: Deep-learning-based batch integration, reference-mapped cell-type annotation, multi-modal CITE-seq / multiome analysis, spatial transcriptomics deconvolution.
Notes
Requires Python with scvi-tools (PyTorch + AnnData stack) installed locally. GPU recommended for larger datasets. Expects a raw-count layer (not log-normalized) when registering AnnData via setup_anndata.
Sources
Installed this tool?
Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.