scvi-tools

Overview

scvi-tools is a comprehensive Python framework for probabilistic models in single-cell genomics. Built on PyTorch and PyTorch Lightning, it provides deep generative models using variational inference for analyzing diverse single-cell data modalities. Current stable release: scvi-tools 1.5.0 (July 2026).

Model namespaces matter: core models (scVI, scANVI, totalVI, MultiVI, PeakVI, AUTOZI, CondSCVI, DestVI, LinearSCVI, AmortizedLDA) live under scvi.model. Most other models (VeloVI, contrastiveVI, CellAssign, PoissonVI, scBasset, MrVI, MethylVI/MethylANVI, CytoVI, SysVI, Decipher, gimVI, scVIVA, ResolVI, Stereoscope, Solo, totalANVI, DIAGVI, DRVI, JointEmbeddingSCVI) live under scvi.external. The reference files specify the correct namespace per model. Class attributes are mostly upper-case (scvi.external.VELOVI, scvi.external.POISSONVI, scvi.external.RNAStereoscope) — take the exact spelling from the reference file rather than from the prose name.

When to Use This Skill

Use this skill when:

• Analyzing single-cell RNA-seq data (dimensionality reduction, batch correction, integration)
• Working with single-cell ATAC-seq or chromatin accessibility data
• Integrating multimodal data (CITE-seq, multiome, paired/unpaired datasets)
• Analyzing spatial transcriptomics data (deconvolution, spatial mapping)
• Performing differential expression analysis on single-cell data
• Conducting cell type annotation or transfer learning tasks
• Working with specialized single-cell modalities (methylation, cytometry, RNA velocity)
• Building custom probabilistic models for single-cell analysis

Core Capabilities

scvi-tools provides models organized by data modality:

1. Single-Cell RNA-seq Analysis

Core models for expression analysis, batch correction, and integration. See references/models-scrna-seq.md for:

• scVI: Unsupervised dimensionality reduction and batch correction
• scANVI: Semi-supervised cell type annotation and integration
• AUTOZI: Zero-inflation detection and modeling
• VeloVI: RNA velocity analysis
• contrastiveVI: Perturbation effect isolation
• DRVI: Unsupervised disentangled representation learning — a decoder-side constraint keeps latent dimensions from mixing, so they can be read one at a time
• JointEmbeddingSCVI: scVI variant trained with a cross-correlation objective on a binomially thinned view of the counts, for embeddings robust to dropout

2. Chromatin Accessibility (ATAC-seq)

Models for analyzing single-cell chromatin data. See references/models-atac-seq.md for:

• PeakVI: Peak-based ATAC-seq analysis and integration
• PoissonVI: Quantitative fragment count modeling
• scBasset: Deep learning approach with motif analysis

3. Multimodal & Multi-omics Integration

Joint analysis of multiple data types. See references/models-multimodal.md for:

• totalVI: CITE-seq protein and RNA joint modeling
• totalANVI: Semi-supervised CITE-seq (totalVI with cell-type labels)
• MultiVI: Paired and unpaired multi-omic integration (MuData-based)
• MrVI: Multi-resolution cross-sample analysis
• DIAGVI: Diagonal integration of unpaired single-cell datasets (added in 1.4.3)

4. Spatial Transcriptomics

Spatially-resolved transcriptomics analysis. See references/models-spatial.md for:

• DestVI: Multi-resolution spatial deconvolution
• Stereoscope: Cell type deconvolution
• Tangram: Spatial mapping and integration
• scVIVA: Cell-environment relationship analysis

5. Specialized Modalities

Additional specialized analysis tools. See references/models-specialized.md for:

• MethylVI/MethylANVI: Single-cell methylation analysis
• CytoVI: Flow/mass cytometry batch correction
• Solo: Doublet detection
• CellAssign: Marker-based cell type annotation

Typical Workflow

All scvi-tools models follow a consistent API pattern:

# 1. Load and preprocess data (AnnData format)
import scvi
import scanpy as sc

adata = scvi.data.heart_cell_atlas_subsampled()
sc.pp.filter_genes(adata, min_counts=3)
sc.pp.highly_variable_genes(adata, n_top_genes=1200)

# 2. Register data with model (specify layers, covariates)
scvi.model.SCVI.setup_anndata(
    adata,
    layer="counts",  # Use raw counts, not log-normalized
    batch_key="batch",
    categorical_covariate_keys=["donor"],
    continuous_covariate_keys=["percent_mito"]
)

# 3. Create and train model
model = scvi.model.SCVI(adata)
model.train()

# 4. Extract latent representations and normalized values
latent = model.get_latent_representation()
normalized = model.get_normalized_expression(library_size=1e4)

# 5. Store in AnnData for downstream analysis
adata.obsm["X_scVI"] = latent
adata.layers["scvi_normalized"] = normalized

# 6. Downstream analysis with scanpy
sc.pp.neighbors(adata, use_rep="X_scVI")
sc.tl.umap(adata)
sc.tl.leiden(adata)

Key Design Principles:

• Raw counts required: Models expect unnormalized count data for optimal performance
• Unified API: Consistent interface across all models (setup → train → extract)
• AnnData-centric: Seamless integration with the scanpy ecosystem
• GPU acceleration: Automatic utilization of available GPUs
• Batch correction: Handle technical variation through covariate registration

Common Analysis Tasks

Differential Expression

Probabilistic DE analysis using the learned generative models:

de_results = model.differential_expression(
    groupby="cell_type",
    group1="TypeA",
    group2="TypeB",
    mode="change",  # Use composite hypothesis testing
    delta=0.25      # Minimum effect size threshold
)

See references/differential-expression.md for detailed methodology and interpretation.

Model Persistence

Save and load trained models:

# Save model
model.save("./model_directory", overwrite=True)

# Load model
model = scvi.model.SCVI.load("./model_directory", adata=adata)

Batch Correction and Integration

Integrate datasets across batches or studies:

# Register batch information
scvi.model.SCVI.setup_anndata(adata, batch_key="study")

# Model automatically learns batch-corrected representations
model = scvi.model.SCVI(adata)
model.train()
latent = model.get_latent_representation()  # Batch-corrected

Theoretical Foundations

scvi-tools is built on:

• Variational inference: Approximate posterior distributions for scalable Bayesian inference
• Deep generative models: VAE architectures that learn complex data distributions
• Amortized inference: Shared neural networks for efficient learning across cells
• Probabilistic modeling: Principled uncertainty quantification and statistical testing

See references/theoretical-foundations.md for detailed background on the mathematical framework.

Additional Resources

• Workflows: references/workflows.md contains common workflows, best practices, hyperparameter tuning, and GPU optimization
• Model References: Detailed documentation for each model category in the references/ directory
• Official Documentation: https://docs.scvi-tools.org/en/stable/
• Tutorials: https://docs.scvi-tools.org/en/stable/tutorials/index.html
• API Reference: https://docs.scvi-tools.org/en/stable/api/index.html

Installation

Requires Python 3.12+ (scvi-tools 1.4 dropped older versions).

uv pip install scvi-tools
# For GPU support
uv pip install "scvi-tools[cuda]"

For reproducible environments, pin a version: uv pip install scvi-tools==1.5.0.post1.

The scanpy steps in the workflow above need one extra: sc.tl.leiden raises ModuleNotFoundError on a stock install, so add uv pip install "scanpy[leiden]" (or leidenalg) before clustering.

Compute backends: training runs on PyTorch (CPU/GPU/TPU). JAX support was removed in 1.5.0 — scvi.model.JaxSCVI no longer exists (AttributeError), and models that once defaulted to JAX now run on PyTorch, including scvi.external.MRVI and scvi.external.Tangram. An experimental MLX backend for Apple silicon (scvi.model.mlxSCVI) remains, and raises ModuleNotFoundError until mlx is installed.

Best Practices

1. Use raw counts: Always provide unnormalized count data to models
2. Filter genes: Remove low-count genes before analysis (e.g., min_counts=3)
3. Register covariates: Include known technical factors (batch, donor, etc.) in setup_anndata
4. Feature selection: Use highly variable genes for improved performance
5. Model saving: Always save trained models to avoid retraining
6. GPU usage: Enable GPU acceleration for large datasets (accelerator="gpu")
7. Scanpy integration: Store outputs in AnnData objects for downstream analysis
8. Out-of-core training: For collections too large to hold in memory, scvi.dataloaders.AnnbatchDataModule (1.5.0+) wraps an annbatch.Loader over sharded Zarr and passes batch and covariate keys through to the model

Try it

A self-contained check that this skill still works. No account, no key, no download, no GPU.

Data — generated inline by the library itself, which is why the frontmatter declares datasets: []. scvi.data.synthetic_iid() ships with scvi-tools and draws 400 cells by 100 genes across two batches and three labels. The real-data route is the Typical Workflow block above, which pulls the heart cell atlas from a server; this section deliberately takes the other path, because every claim below is about how a trained model behaves and a fetched matrix would put somebody else's host in the way of testing that. It also draws every label from the same distribution, so nothing in it is truly differentially expressed — which is what makes the differential-expression result below a test rather than a decoration.

Run — needs Python 3.12+. Takes well under a minute on CPU:

uv pip install "scvi-tools==1.5.0.post1"

import numpy as np
import scvi

scvi.settings.seed = 0                      # every number below is reproducible with this
adata = scvi.data.synthetic_iid()           # 400 cells x 100 genes, 2 batches, 3 labels
assert adata.shape == (400, 100)

# The trap the Overview opens with: the prose name is not the class attribute, and the
# namespace differs per model. CytoVI is spelled CYTOVI, and it is in scvi.external.
assert hasattr(scvi.external, "CYTOVI")
assert not hasattr(scvi.external, "CytoVI")
assert not hasattr(scvi.model, "CYTOVI")

# JAX went away in 1.5.0. This is an AttributeError, not a missing optional dependency.
try:
    scvi.model.JaxSCVI
    raise SystemExit("JaxSCVI still exists — this skill is out of date")
except AttributeError:
    print("JaxSCVI        : AttributeError, as documented")

scvi.model.SCVI.setup_anndata(adata, batch_key="batch", labels_key="labels")
model = scvi.model.SCVI(adata, n_latent=10)
model.train(max_epochs=20)

# One row per cell, one column per latent dimension.
latent = model.get_latent_representation()
assert latent.shape == (adata.n_obs, 10)

# library_size is honoured: each cell's normalized profile sums to it.
norm = model.get_normalized_expression(library_size=1e4)
assert np.allclose(np.asarray(norm).sum(axis=1), 1e4, rtol=1e-5)

# One row per gene, carrying the composite-hypothesis columns `mode="change"` adds.
de = model.differential_expression(
    groupby="labels", group1="label_0", group2="label_1",
    mode="change", delta=0.25, silent=True,
)
assert len(de) == adata.n_vars
assert {"proba_de", "lfc_mean", "is_de_fdr_0.05"} <= set(de.columns)

# A reloaded model is the same model, not merely a similar one.
model.save("scvi_try_it", overwrite=True)
reloaded = scvi.model.SCVI.load("scvi_try_it", adata=adata)
assert np.array_equal(reloaded.get_latent_representation(), latent)

print("latent shape   :", latent.shape)
print("row sums       : %.2f" % np.asarray(norm).sum(axis=1).mean())
print("genes called DE:", int(de["is_de_fdr_0.05"].sum()), "of", len(de))
print("max proba_de   : %.4f" % de["proba_de"].max())

Expect

Invariants — these hold across versions, and a failure means the skill is wrong:

• latent is (400, 10) — one row per cell, one column per latent dimension.
• Every row of get_normalized_expression(library_size=1e4) sums to 10000.
• differential_expression returns one row per gene (100), carrying proba_de, lfc_mean and is_de_fdr_0.05.
• A saved model reloads to identical latent coordinates, not approximately equal ones.
• scvi.external.CYTOVI exists while scvi.external.CytoVI does not, and scvi.model.JaxSCVI raises AttributeError.

Observed on scvi-tools 1.5.0.post1 / torch 2.13.0 / Python 3.12.3, 2026-08-30 — a mismatch here is drift to investigate, not a bug:

JaxSCVI        : AttributeError, as documented
latent shape   : (400, 10)
row sums       : 10000.00
genes called DE: 0 of 100
max proba_de   : 0.2926

genes called DE: 0 is the line to read closely. The generator draws all three labels from one distribution, so a correct change-mode test should call nothing significant on it; a non-zero count means the test is finding structure that is not there. With the seed pinned the run is deterministic — repeated runs returned these numbers unchanged.