single-cell-multi-omics-analysis-scvi

Batch correction for single-cell multi-omics data is performed using probabilistic models from scvi-tools, which take raw counts in an AnnData object and produce a batch-corrected latent representation for joint analysis.

Spatial transcriptomics and CITE-seq analysis are supported through specialized model families like totalVI, enabling integrated processing of RNA, chromatin accessibility, and protein measurements within a unified API workflow.

Cell type annotation in multimodal single-cell datasets is handled by the scANVI probabilistic model, which enables semi-supervised classification by leveraging labeled data during the integration workflow.

GPU acceleration for scalable probabilistic modeling of scRNA-seq data is supported via PyTorch Lightning, enabling faster training of scvi-tools models on large-scale multi-omics datasets.

Minimal preprocessing is required for single-cell ATAC-seq data because the AnnData-centric workflow accepts raw counts directly, allowing probabilistic models to perform integration and batch correction without extensive prior normalization.