# Analyze Single-Cell Omics With scvi-tools

Single-cell omics projects need careful model setup, batch handling, and uncertainty-aware interpretation. This skill guides scvi-tools workflows for scVI, totalVI, MultiVI, differential expression, and spatial analysis.

## Install

```bash
npx skillstore add k-dense-ai/scvi-tools
```

## Metadata

- Status: approved
- Slug: k-dense-ai-scvi-tools
- Skillstore revision: r1
- Version status: missing
- Tree hash: e25c60baef2f8fa803cce581bb473df7b098911f286567049f4285db552f5c3a
- Author: K-Dense-AI
- GitHub username: K-Dense-AI
- License: BSD-3-Clause license
- Repository: https://github.com/K-Dense-AI/claude-scientific-skills/tree/main/scientific-skills/scvi-tools
- Ref: b8ca75d2c0a7e7102978993058777d82b8ab2610
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 73
- Public page: https://skillstore.pages.dev/skills/k-dense-ai-scvi-tools
- Manifest: https://skillstore.pages.dev/api/skills/k-dense-ai-scvi-tools/manifest

## Capabilities

- Explains when to use scVI, scANVI, totalVI, MultiVI, PeakVI, DestVI, and related models.
- Guides AnnData setup with raw counts, layers, batch keys, and technical covariates.
- Outlines training, latent representation extraction, normalization, and Scanpy downstream analysis.
- Describes probabilistic differential expression with effect sizes, uncertainty, and composite hypotheses.
- Covers model saving, transfer learning, GPU usage, and workflow best practices.

## Use Cases

- Correct batch effects across studies: Plan an scVI workflow that registers batch covariates and produces integrated latent representations for downstream clustering.
- Compare cell states with uncertainty: Use probabilistic differential expression guidance to compare groups while considering effect size and posterior uncertainty.
- Integrate RNA, protein, and accessibility data: Select totalVI, MultiVI, or PeakVI workflows for multimodal experiments and compatible single-cell data structures.

## Prompt Templates

### Choose a starter model

```
Help me choose the right scvi-tools model for my single-cell dataset. Ask about modality, batches, labels, and analysis goals.
```

### Prepare AnnData for scVI

```
Review my planned AnnData fields for scVI. Check raw counts, batch keys, layers, covariates, and feature selection choices.
```

### Plan differential expression

```
Design a scvi-tools differential expression workflow for two cell groups. Include effect size thresholding and interpretation guidance.
```

### Design a multimodal integration workflow

```
Create an advanced scvi-tools plan for integrating paired and unpaired multi-omics data with batch covariates and validation checks.
```

## Limitations

- It provides guidance and examples, but it does not run analysis by itself.
- Users must supply valid AnnData objects with raw count data.
- Large datasets may need GPU resources and careful memory planning.
- Biological interpretation still requires domain review and experimental context.

## Best Practices

- Start with raw count data and keep normalized values separate from model inputs.
- Register known technical covariates before training each model.
- Save trained models and document hyperparameters for reproducible analysis.

## Anti Patterns

- Do not train scvi-tools models on log-normalized expression values.
- Do not ignore batch, donor, or chemistry effects when they are known.
- Do not treat model output as final biological proof without validation.

## Security Audit

- Audited at: 2026-07-06T19:07:44.569\+00:00
- Summary: No executable malware, command injection, or data exfiltration was found. The static command, network, and reconnaissance findings are documentation false positives from markdown formatting, examples, formulas, and official links. A low-severity semantic issue remains because the skill steers users toward the author's hosted platform for complex workflows.

## Stats

- Views: 262
- Downloads: 7
- Favorites: 0
- Popularity score: 0