scvi-tools

Welcome to the scvi-tools organization. We provide state-of-the-art probabilistic models tailored for analyzing single-cell omics data. Those enable researchers to gain biological insights with scalable algorithms.
These models provide a consistent API making it easy to integrate it into your current analysis pipeline. scvi-tools is part of scverse.
This is an open science initiative, please contribute your own models to allow the single-cell community to leverage your reference datasets. Learn how to upload your model in our HubModel tutorials.

Model Overview

scvi-tools offers a comprehensive suite of models designed to address various challenges in single-cell data analysis.

Current HubModels

• scVI:
  • A variational autoencoder for dimensionality reduction, batch correction, and clustering.
  • See all models in scVI collection
• scANVI:
  • A semi-supervised model designed for label prediction, especially in cases of partially labeled data.
  • See all models at scANVI collection
• totalVI:
  • A multi-modal model for joint analysis of RNA and protein data, additionally allowing imputation of missing protein data.
  • See all models in totalVI collection
• DestVI:
  • A deconvolution model for prediction of single-cell profiles given subcellular spatial transcriptomics data. We provide here pre-trained single cell models (CondSCVI).
  • See all models in DestVI collection
• Stereoscope:
  • A deconvolution model for prediction of cell-type composition given subcellular spatial transcriptomics data. We provide here pre-trained single cell models.
  • See all models in Stereoscope collection

Explore the full list of models in scvi-tools in our user guide.
Please reach out on discourse, if you want to add additional models to HuggingFace.

Key Applications

These models have been applied to a wide array of biological questions, such as:

• Batch correction across experiments.
• Identification of rare cell populations.
• Multi-modal integration of single-cell RNA, and protein data.
• Differential expression and abundance analysis in disease contexts.

For hands-on examples, refer to our tutorials.
Learn how to apply scvi-hub for analysis of query datasets in our HLCA tutorial.
Discover how to efficiently access CELLxGENE census using our minified models in our CELLxGENE census tutorial.

Publications

• Original scvi-tools Paper:
  • Published in Nature Biotechnology, this paper introduces the foundational principles of scvi-tools.
• scvi-hub Preprint:
  • This manuscript showcases real-world applications of scvi-hub in diverse biological contexts and provides building blocks
  • to apply these models in your own research

How to Get Started

1. Visit our official documentation to get started with installation and explore our API.
2. Follow our tutorials for step-by-step guides on using scvi-tools effectively.
3. Dive into our models to see how to apply them to your single-cell analysis.

Contact

• Website: https://scvi-tools.org
• GitHub: https://github.com/scverse/scvi-tools
• Tutorials: scvi-tools Tutorials
• User questions: Discourse