scvi-mrvi

What problem does it solve?

MRVI enables multi-donor single-cell RNA-seq analyses by modeling sample identity as biological variation, providing cell-type-specific insights and covariate-aware differential expression.

Core Features & Use Cases

• Dual latent representations: a sample-invariant space for cell-type clustering and a sample-aware space for studying sample effects.
• Covariate-linked differential expression and differential abundance to attribute changes to covariates.
• Handles cohorts with many donors to compute cell-type-specific sample distances and study heterogeneity.

Quick Start

Setup an AnnData object with sample_id as the biological key, initialize MRVI, and train to obtain both latent representations for downstream analyses.