What data format is required for this skill?

This skill works with AnnData (.h5ad) files and requires raw integer counts in a specific layer to satisfy the requirements of scvi-tools probabilistic models.

Does this skill support reference mapping?

Yes, it provides detailed workflows for using scArches and scANVI to map query datasets to existing reference models for rapid label transfer.

Can it handle spatial transcriptomics?

Absolutely. It includes specific guidance for using DestVI to perform cell type deconvolution when you have a matching scRNA-seq reference.

Which model should I use for CITE-seq data?

For multi-modal data involving both RNA and protein (CITE-seq), the skill recommends using the totalVI model for joint integration and protein denoising.

Single-Cell Deep Learning (scvi-tools)

Name: Single-Cell Deep Learning (scvi-tools)
Author: anthropics

byanthropics

•

147

데이터 과학 및 ML

Streamlines single-cell genomics workflows using scvi-tools for probabilistic data integration and deep learning analysis.

소개

This skill integrates the scvi-tools framework into Claude Code, providing a comprehensive toolkit for deep learning-based single-cell analysis. It assists researchers and bioinformaticians in selecting and implementing the correct probabilistic models—such as scVI for integration, totalVI for CITE-seq, or PeakVI for ATAC-seq—and provides standardized scripts for data preparation, model training, and downstream visualization. By following domain-specific best practices, it ensures proper handling of raw counts, batch effects, and latent space representations within the scverse ecosystem.

주요 기능

147 GitHub stars
Support for spatial transcriptomics deconvolution and RNA velocity
Standardized scripts for data QC, HVG selection, and AnnData validation
Reference mapping and label transfer using scArches and scANVI
Advanced batch correction and multi-dataset integration workflows
Guided model selection for scRNA-seq, scATAC-seq, and multi-modal data

사용 사례

Mapping new query samples to a curated reference atlas for automated cell-type annotation
Jointly analyzing gene expression and protein abundance in CITE-seq experiments
Integrating heterogeneous scRNA-seq datasets to remove technical batch effects

소개

주요 기능

147 GitHub stars
Support for spatial transcriptomics deconvolution and RNA velocity
Standardized scripts for data QC, HVG selection, and AnnData validation
Reference mapping and label transfer using scArches and scANVI
Advanced batch correction and multi-dataset integration workflows
Guided model selection for scRNA-seq, scATAC-seq, and multi-modal data

사용 사례

Mapping new query samples to a curated reference atlas for automated cell-type annotation
Jointly analyzing gene expression and protein abundance in CITE-seq experiments
Integrating heterogeneous scRNA-seq datasets to remove technical batch effects