scvi-tools: a library for deep probabilistic analysis of single-cell omics data

Although isoform diversity is acknowledged as a fundamental and pervasive aspect of gene expression in higher eukaryotes, it is often omitted from single-cell studies due to quantification challenges inherent to commonly used short-read sequencing technologies. To address this issue, we have developed a suite of computational tools to investigate isoform variation by focusing on splice junction usage patterns, which can often be well characterized in spite of technical difficulties. Our method, which we name scQuint (Single-Cell QUantification of INTrons), can perform accurate quantification, dimensionality reduction, and differential splicing analysis using short-read, full-length single-cell RNA-seq data. Notably, scQuint does not require transcriptome annotations and is robust to technical artifacts. In applications across diverse mouse tissues from Tabula Muris and the primary motor cortex from the BRAIN Initiative Cell Census Network, we find evidence of strong cell-type-specific isoform variation, complementary to total gene expression, and also identify a large volume of previously unannotated splice junctions. As a community resource, we provide ways to interactively visualize and explore these results, accessible at https://github.com/songlab-cal/scquint-analysis/ .

show abstract

“…The VAE was implemented using PyTorch [60] and scvi-tools [61]. The following is the generative model, repeated for each cell:…”

Section: Methodsmentioning

confidence: 99%

Robust and annotation-free analysis of alternative splicing across diverse cell types in mice

Benegas

Fischer

Song

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…For example, the user can select a group according to a combination of cell type, sample, tissue and experimental group. DE is computed using the scVI model [4] from scvi-tools [3], which enables quick computation even when using only CPUs. The results are displayed in the form of an interactive volcano plot (log fold change vs p-value) and MA plot (log fold change vs mean expression) that display gene descriptions upon mouseover, and sortable tabular results that can be downloaded in csv and Excel format.…”

Section: Overview Of the Scdefg And Wormcells-viz Toolsmentioning

confidence: 99%

“…The scvi-tools framework offers several models for single cell omics data, and for scRNA-seq in particular offers the scVI model [4], which is bayesian hierachical generative model that leverages variational autoencoders to enable robust statistical analysis. It is built with PyTorch (pytorch.org) and has been extensively validated [3]. Training the scVI model only requires a gene count matrix, as outputted by scRNA-seq alignment software such as Cell Ranger [5].…”

Section: Rationale For Using Scvi-toolsmentioning

confidence: 99%

“…The scvi-tools codebase (github.com/YosefLab/scvi-tools) was first introduced in 2017, and published in 2018 [4], with consistent updates and improvements since [3]. It now boasts a mature and professional API and codebase that follows industry best practices, and has over 37 unique contributors and 56 releases.…”

Section: Consistent Development and Contributorsmentioning

confidence: 99%

“…In this spirit we present two tools: scdefg , for performing inter-active differential expression (DE), and wormcells-viz , for visualization of gene expression data. These tools leverage the anndata file format [2], a standard file format for scRNA-seq data and annotated data matrices (anndata.readthedocs.io) and scvi-tools [3], a popular framework for generative modeling of scRNA-seq data and statistical analysis (scvi-tools.org).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Single cell tools for WormBase

Beltrame

Arnaboldi

Sternberg

2021

Preprint

Self Cite

View full text Add to dashboard Cite

We present two web apps for interactively performing common tasks with single cell RNA sequencing data: scdefg for interactive differential expression and wormcells-viz for visualization of gene expression. We made these tools available with public C. elegans datasets curated by WormBase at https://single-cell.wormbase.org. They can also be readily deployed for use with any other datasets using the source code available at https://github.com/WormBase/scdefg/ and at https://github.com/WormBase/wormcells-viz.

show abstract

Interpretable generative deep learning: an illustration with single cell gene expression data

2022

View full text Add to dashboard Cite

Deep generative models can learn the underlying structure, such as pathways or gene programs, from omics data. We provide an introduction as well as an overview of such techniques, specifically illustrating their use with single-cell gene expression data. For example, the low dimensional latent representations offered by various approaches, such as variational auto-encoders, are useful to get a better understanding of the relations between observed gene expressions and experimental factors or phenotypes. Furthermore, by providing a generative model for the latent and observed variables, deep generative models can generate synthetic observations, which allow us to assess the uncertainty in the learned representations. While deep generative models are useful to learn the structure of high-dimensional omics data by efficiently capturing non-linear dependencies between genes, they are sometimes difficult to interpret due to their neural network building blocks. More precisely, to understand the relationship between learned latent variables and observed variables, e.g., gene transcript abundances and external phenotypes, is difficult. Therefore, we also illustrate current approaches that allow us to infer the relationship between learned latent variables and observed variables as well as external phenotypes. Thereby, we render deep learning approaches more interpretable. In an application with single-cell gene expression data, we demonstrate the utility of the discussed methods.

show abstract

scvi-tools: a library for deep probabilistic analysis of single-cell omics data

Cited by 61 publications

References 83 publications

Robust and annotation-free analysis of alternative splicing across diverse cell types in mice

Robust and annotation-free analysis of alternative splicing across diverse cell types in mice

Single cell tools for WormBase

Interpretable generative deep learning: an illustration with single cell gene expression data

Contact Info

Product

Resources

About