scvi-tools

Correct batch effects in multimodal single-cell datasets by applying probabilistic models that learn robust latent representations. This approach integrates scRNA-seq, scATAC-seq, and protein data while discriminating biological signal from technical noise across different batches.

Perform Bayesian differential expression on scRNA-seq data by using probabilistic models with a batch-aware decoder. This method enables end-to-end analysis by calculating expression differences while accounting for technical noise and batch variations.

You can integrate scATAC-seq and scRNA-seq modalities for joint analysis using cross-modality imputation. Probabilistic modeling supports scalable amortized inference to align RNA, ATAC, and protein data into a shared latent space.

Probabilistic modeling handles large-scale single-cell integration through scalable amortized inference. It processes raw-count inputs to generate batch-corrected latent representations, supporting cross-study meta-analysis and large datasets efficiently.

Raw count inputs are required for single-cell batch correction and imputation using this probabilistic workflow. The models process raw counts to learn latent representations, perform imputation, and execute differential expression with a consistent API.

Use latent-space interpretation for single-cell clustering and visualization to separate biological signal from technical noise. Probabilistic models generate batch-corrected latent representations that improve downstream clustering accuracy and visualization quality.