scvi-tools

To integrate scRNA-seq datasets and perform batch correction, you can use deep generative models like scVI to remove technical variations across experiments. This process reveals true biological signals by aligning heterogeneous AnnData objects into a unified, corrected latent space.

Cell type annotation in single-cell genomics is best handled using scANVI, which transfers labels from reference atlases or performs de novo annotation. This deep learning method leverages semi-supervised learning to accurately classify cell identities across new scRNA-seq samples.

You can estimate RNA velocity to model cell differentiation dynamics by applying the veloVI deep generative model. This computes the splicing kinetics of RNA within your single-cell genomics dataset, mapping the directional trajectory of cell state transitions.

This single-cell analysis approach supports multi-modal data integration using specialized models like totalVI, PeakVI, and MultiVI. These frameworks jointly analyze RNA, ATAC, and protein data within MuData objects to provide a comprehensive view of cellular states.

You need a Python environment with PyTorch, scanpy, and scvi-tools installed to run this analysis. These dependencies are essential for processing AnnData structures, executing deep generative models, and managing tensor computations for single-cell genomics workflows.