When should I use scvi-tools instead of Scanpy?

Use scvi-tools when you need advanced probabilistic modeling, such as batch correction for complex datasets, multimodal integration, or uncertainty-aware differential expression. Scanpy is better suited for standard preprocessing, basic clustering, and visualization.

What data format does scvi-tools require?

scvi-tools is designed to work with AnnData objects. Crucially, the models require raw, unnormalized count data as input to correctly model the underlying data distribution.

Can I use this for spatial transcriptomics?

Absolutely. This skill supports specialized spatial models like DestVI for deconvolution, Stereoscope for cell type mapping, and Tangram for integrating spatial data with single-cell references.

Does this skill support GPU acceleration?

Yes, scvi-tools is built on PyTorch and PyTorch Lightning, meaning it can automatically utilize available GPUs to significantly speed up the training of deep generative models on large datasets.

scvi-tools: Probabilistic Single-Cell Analysis

Name: scvi-tools: Probabilistic Single-Cell Analysis
Author: Sologa

bySologa

Data Science & ML

Implement deep generative models for single-cell omics analysis, enabling batch correction, multimodal integration, and probabilistic differential expression.

About

This skill integrates scvi-tools into Claude Code, providing a specialized framework for analyzing single-cell genomic data using deep generative models. Built on PyTorch, it excels at complex tasks like batch correction (scVI), transfer learning (scANVI), and multimodal integration (TOTALVI, MultiVI). It is particularly useful for researchers and bioinformaticians who need to handle technical variation across datasets, perform uncertainty-aware differential expression, or analyze paired multi-omic data with high statistical precision beyond standard analysis pipelines.

Key Features

Multimodal data integration for CITE-seq, Multiome, and spatial transcriptomics
Semi-supervised cell type annotation and transfer learning with scANVI
Uncertainty-aware differential expression analysis using Bayesian inference
0 GitHub stars
Native integration with the AnnData and Scanpy ecosystem for seamless workflows
Probabilistic batch correction and dimensionality reduction using scVI

Use Cases

Jointly modeling protein and RNA expression from CITE-seq data to identify rare cell states
Integrating heterogeneous single-cell RNA-seq datasets across different labs or technologies
Deconvolving spatial transcriptomics data to map cell types across complex tissue sections

About

Key Features

Multimodal data integration for CITE-seq, Multiome, and spatial transcriptomics
Semi-supervised cell type annotation and transfer learning with scANVI
Uncertainty-aware differential expression analysis using Bayesian inference
0 GitHub stars
Native integration with the AnnData and Scanpy ecosystem for seamless workflows
Probabilistic batch correction and dimensionality reduction using scVI

Use Cases

Jointly modeling protein and RNA expression from CITE-seq data to identify rare cell states
Integrating heterogeneous single-cell RNA-seq datasets across different labs or technologies
Deconvolving spatial transcriptomics data to map cell types across complex tissue sections