scvi-peakvi

You can batch correct scATAC-seq data by loading your AnnData into PeakVI and training the model to generate a probabilistic, batch-corrected latent embedding that accounts for technical variation across experiments.

To analyze differential accessibility, use PeakVI to generate a latent representation of your scATAC-seq data, which enables region-level differential accessibility analysis between distinct cell populations.

Yes, you need a GPU to run scATAC-seq clustering with PeakVI. The model relies on scvi-tools and PyTorch APIs, which require GPU acceleration to process the data and train effectively.

Yes, PeakVI works with AnnData and scanpy workflows. You can load your scATAC-seq AnnData object, configure the model with scvi-tools, and integrate the resulting latent embeddings into your existing analysis pipeline.

PeakVI handles large-scale scATAC-seq integration by applying a probabilistic latent space that scales to over one million cells, allowing you to cluster datasets and compare chromatin accessibility without being limited by batch effects.

scATAC-seq clustering often shows technical variation due to batch effects across experiments. PeakVI addresses this by generating a batch-corrected latent space that accounts for this technical variation during model training.