scvi-tools

Batch correction on single-cell RNA-seq data is performed using deep generative models like scVI, which learn latent representations to remove technical variation while preserving biological diversity in AnnData objects.

Yes, you can jointly analyze RNA and protein expression from CITE-seq datasets using the totalVI model, which integrates multimodal data to simultaneously identify cell types and perform differential expression across modalities.

Zero-inflation in single-cell omics data is handled natively by probabilistic models like scVI, which use deep generative architectures to model technical dropout effects during dimensionality reduction and latent representation extraction.

Yes, scvi-tools works directly with AnnData objects for multimodal data integration, supporting specialized models like MultiVI and totalVI to jointly process RNA-seq, ATAC-seq, and protein expression modalities.

Cell type annotation in single-cell genomics is best achieved using the scANVI model, a semi-supervised deep learning approach that leverages probabilistic latent representations to accurately classify and annotate cell populations.

Yes, PyTorch is required as the underlying framework for trajectory inference, as the probabilistic models are built upon PyTorch Lightning to train deep generative networks for analyzing cellular dynamics.