scviR

scviR defines interfaces from R to scvi-tools for probabilistic single-cell data analysis using variational autoencoders.

Key Features:

• totalVI integration: Provides R interfaces to run totalVI for analyzing CITE-seq data combining transcriptome and antibody-derived tag measurements.
• OSCA workflow comparison: Enables comparison of standard single-cell analysis outputs with totalVI-based quantifications through vignettes aligned with Chapter 12 of the OSCA book.
• Variational autoencoder framework: Supports building understanding of probabilistic methods based on variational autoencoders for single-cell data analysis.

Scientific Applications:

• CITE-seq analysis: Facilitates joint analysis of gene expression and surface protein abundance data from CITE-seq experiments through totalVI.
• Benchmarking and validation: Allows direct comparison between conventional single-cell analysis pipelines and deep generative model-based quantifications.

Methodology:

The package implements interfaces from R to scvi-tools, with current functionality focused on totalVI for variational autoencoder-based analysis of CITE-seq data.