scvi-tools

Correct batch effects in scRNA-seq data by applying probabilistic models like scVI to AnnData objects containing raw counts, which denoise technical variations and infer consistent biological latent representations across samples.

Multimodal integration of scRNA-seq and scATAC-seq is supported through models like MultiVI and PeakVI, allowing you to jointly analyze RNA and chromatin accessibility within a single probabilistic framework to map cross-modality relationships.

Differential expression on spatial transcriptomics data is handled by training probabilistic generative models on AnnData inputs, which denoise the spatial measurements and enable reliable identification of differentially expressed genes across spatial regions.

GPU acceleration is not strictly required but is leveraged through PyTorch to optimize training speeds for probabilistic models on AnnData objects, significantly improving performance when processing large-scale single-cell datasets.

Use scANVI instead of scVI when you have partial cell type labels available, as scANVI extends the scVI framework to incorporate supervised annotation, enabling you to transfer labels to unannotated cells within your dataset.

Deconvolve complex tissues using CITE-seq data by applying the TOTALVI model to jointly model RNA and protein measurements, yielding denoised expressions and latent representations that resolve mixed cell populations.