What problem does it solve?
This Skill solves the challenge of performing high-quality, deep-learning single-cell analysis—especially integration, denoising, multimodal learning, and reference-based mapping—without stitching together many manual steps.
Core Features & Use Cases
- Model selection for common scvi-tools tasks: Choose the right model for scRNA-seq, scATAC-seq, CITE-seq, multiome (RNA+ATAC), spatial deconvolution, velocity, and system-level batch correction.
- End-to-end workflow guidance: Covers environment setup, data preparation, training, latent embedding extraction, clustering/visualization, and model-based differential analysis.
- Label transfer and reference mapping: Use scANVI/scArches to map query cells into a reference atlas for consistent annotation across studies.
Example use cases
- You need batch-corrected scRNA-seq integration: start from scVI/scANVI (depending on whether labels exist).
- You have CITE-seq (RNA + protein): use totalVI to jointly model modalities and denoise protein measurements.
- You want label transfer to annotate a new dataset: use scANVI/scArches to map query data to a pretrained reference.
Quick Start
Tell your assistant: “Use the scvi-tools skill to run a complete scRNA-seq integration workflow with scVI on my AnnData file and then output the integrated latent space for clustering.”
