chevreul
Chevreul
This package includes a set of Shiny apps for interactive exploration of single cell RNA sequencing (scRNA-seq) datasets processed as SingleCellExperiments
A demo with a developing human retina scRNA-seq dataset from Shayler et al. is available here
Chevreul includes tools for:
- Interactive data analysis and visualization
- Louvain clustering at a range of resolutions
- Dimensionality reduction of raw sequencing data.
- Integration (batch correction) of multiple datasets
[!WARNING] Chevreul was designed for full-length smart-seq based single cell data. Default settings may not be appropriate for droplet (10x) data, though most can be adjusted. Keep in mind best practices regarding normalization, dimensional reduction, etc. when using.
Installation instructions
Chevreul depends on a minimum R version 4.4. The current
Chevreul loads three Bioconductor packages
These enable standardized processing, plotting, and interactive analysis of SingleCellExperiments, respectively.
Chevreul depends on a minimum R version >=4.4 Get the latest stable
R release from CRAN. Then install
Chevreul and its dependencies using the following code:
install.packages("remotes")
install.packages("BiocManager")
BiocManager::install("cobriniklab/chevreul")
Troubleshooting installation
Dependency management
When installing an R package like Chevreul with many dependencies, conflicts with existing installations can arise. This is a common issue in R package management. Here are some strategies to address this problem:
-
Consider renv for dependency management. This tool creates isolated environments for each project, ensuring that package versions don’t conflict across different projects.
-
Use the conflicted Package The conflicted package provides an alternative conflict resolution strategy. It makes every conflict an error, forcing you to choose which function to use
Slow internet connection
When installing R packages on slow internet connections, several issues can arise, particularly with larger packages or when using functions like remotes::install_github(). Here are some strategies to address bandwidth-related problems:
Set a longer timeout for downloads: options(timeout = 9999999)
Specify the download method: options(download.file.method = "libcurl")
Quick Start
After successful installation of the Chevreul package from Bioconductor, load the package along with the SingleCellExperiment example data provided with the package.
# Load packages
library(chevreul)
# Load and view example dataset
data("small_example_dataset")
small_example_dataset
Run clustering on a single object
The sce_clustering_workflow() function performs key processing steps,
including quality control filtering, normalization and log
transformation, dimensionality reduction, Louvain clustering across
various resolutions, and the identification of cluster-specific marker
genes or transcripts.
data("medium_example_dataset")
medium_example_dataset_processed <- sce_clustering_workflow(medium_example_dataset, resolution = 0.6, experiment_name = "sce", organism = "human")
Get a first look at a processed dataset using an interactive shiny app
The function minimalChevreulApp opens an interactive shiny application
that can be used for visualization and exploration of the processed
dataset.
minimalChevreulApp(tiny_sce)
Transcript-level quantification
For transcript-level analysis, users must incorporate transcript-level data into the SingleCellExperiment object as an alternative experiment before initiating the Chevreul processing pipeline. This step is crucial for enabling detailed exploration at the transcript level.
Transcripts may be quantified using any of several available methods, including alignment-free methods best used with well-annotated transcriptomes (Salmon, Kallisto), alignment-based methods best used to detect novel isoforms (StringTie2), or long-read methods for use with long-read sequencing data (IsoQuant).
Integration implementation
The sce_integrate() function in Chevreul implements integration (batch
correction) of scRNA-seq datasets by using the
batchelor
package.
It accepts a list of SingleCellExperiment objects as input for
integration and stores the corresponding batch information in a metadata
field named ‘batch’. By default, it employs batchelor’s
correctExperiments function to preserve pre-existing data structures
and metadata from input SingleCellExperiment objects within the
integrated output.
Hardware requirements
Recommended minimum hardware requirements for running Chevreul are as follows:
- RAM: A minimum of 16 GB RAM is recommended for initial analysis. However, for larger datasets or more complex analyses, 64 GB or more is advisable.
- CPU: Having multiple cores can be beneficial for parallel processing.
- Storage: Sufficient storage space is necessary, especially for temporary files. The exact amount depends on the size of your datasets
- R Version: Chevruel requires R version 4.4 or greater
It’s important to note that these requirements can vary based on the size and complexity of your dataset. As the number of cells increases, so do the hardware requirements. For instance: A dataset with around 8,000 cells can be analyzed with 8 GB of RAM. For larger datasets or more complex analyses, 64-128 GB of RAM can be beneficial.
Learn More
To learn more about the usage of Bioconductor tools for single-cell RNA-seq analysis. Consult the book Orchestrating Single-Cell Analysis with Bioconductor. The book walks through common workflows for the analysis of single-cell RNA-seq data (scRNA-seq). This book will show you how to make use of cutting-edge Bioconductor tools to process, analyze, visualize, and explore scRNA-seq data