scvi-tools

SUMMARY

What it does

This Agent Skill provides procedural knowledge for using the scvi-tools Python framework, which implements deep generative models (variational autoencoders) for single-cell genomics. The skill covers model selection across data modalities (RNA-seq, ATAC-seq, multimodal, spatial, and specialized modalities), typical workflows (data registration, training, extraction of latent representations), differential expression analysis, model persistence, and batch correction. It also includes installation guidance, best practices, and pointers to reference documentation. The skill is designed for research agents performing advanced single-cell analysis and assumes familiarity with AnnData and scanpy.