PRECAST

PRECAST: a probabilistic embedding and clustering with alignment for spatial transcriptomics data integration.

PRECAST is a package for integrating and analyzing multiple spatially resolved transcriptomics (SRT) datasets, developed by the Jin Liu’s lab. It unifies spatial factor analysis simultaneously with spatial clustering and embedding alignment, requiring only partially shared cell/domain clusters across datasets.

Check out our bioRxiv paper and our Package Website for a more complete description of the methods and analyses.

PRECAST can be used to compare and contrast experimental datasets in a variety of contexts, for instance:

• Across experimental batches
• Across individuals
• Across different conditions (i.e., case and control)
• Across datasets with only partially shared cell/domain clusters

Once multiple datasets are integrated, the package provides functionality for further data exploration, analysis, and visualization. Users can:

• Identify clusters using all data information
• Extract aligned low-dimensional embeddings across datasets
• Recover comparable gene expression matrices among datasets
• Find significant shared (and dataset-specific) gene markers
• Conditional spatially variational genes analysis
• Compare clusters with previously identified domain/cell types
• Visuzlize extracted embeddings using 3-dim tSNE and UMAP
• Visualize clusters and gene expression using tSNE and UMAP

Installation

“PRECAST” depends on the ‘Rcpp’ and ‘RcppArmadillo’ package, which requires appropriate setup of computer. For the users that have set up system properly for compiling C++ files, the following installation command will work.

# Method 1: install PRECAST from CRAN
install.packages('PRECAST')


# For the newest version of PRECAST, users can use method 2 for installation.
# Method 2: Install PRECAST from Github
if (!require("remotes", quietly = TRUE))
    install.packages("remotes")
remotes::install_github("feiyoung/PRECAST")

# If some dependent packages (such as `scater`) on Bioconductor can not be installed nomrally, use following commands, then run abouve command.
if (!require("BiocManager", quietly = TRUE)) ## install BiocManager
    install.packages("BiocManager")
# install the package on Bioconducter
BiocManager::install(c("scater"))

Usage

For usage examples and guided walkthroughs, check the vignettes directory of the repo.

• Single SRT data analysis
• Toy examples for integrating two bathes
• Integration across experimental batches

For the users that don’t have set up system properly, the following setup on different systems can be referred. ## Setup on Windows system First, download Rtools; second, add the Rtools directory to the environment variable. Users can follow here to add Windows PATH Environment Variable.

Setup on MacOS system

First, install Xcode. Installation about Xcode can be referred here.

Second, install “gfortran” for compiling C++ and Fortran at here.

Setup on Linux system

For parallel computation on Linux, users must use the following system command to set the C_stack unlimited in case of the error R Error: C stack usage is too close to the limit.

ulimit -s unlimited

If you use conda environment on Linux system and some dependent packages (such as scater) can not normally installed, you can search R package at anaconda.org website. We take the scater package as example, and its search result is https://anaconda.org/bioconda/bioconductor-scater. Then you can install it in conda environment by following command.

conda install -c bioconda bioconductor-scater

For the user not using conda environment, if dependent packages (such as scater) not normally installed are in Bioconductor, then use the following command to install the dependent packages.

# install BiocManager
if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
# install the package on Bioconducter
BiocManager::install(c("scater"))

If dependent packages (such as DR.SC) not normally installed are in CRAN, then use the following command to install the dependent packages.

# install the package on CRAN
install.packages("DR.SC")

Demonstration

For an example of typical PRECAST usage, please see our Package Website for a demonstration and overview of the functions included in PRECAST.

NEWs

PRECAST version 1.3 (2022-10-05)

• Fix the issue reported by Boyi Guo. Specifically, the assay name “RNA” used in functions CreatePRECASTObject and PRECAST is replaced by the default assay using DefaultAssay function in Seurat.

• Provide more detailed help file for CreatePRECASTObject function. Users can use ?CreatePRECASTObject in Rstudio to access the help file. In detail, seuList is a list with Seurat object as component, and each Seurat object at least includes the raw expression count matrix, and spatial coordinates in metadata for each data batch, where the spatial coordinates information must be saved in the metadata of Seurat, named “row” and “col” for each data batch. See the help file for more details.

• Add the data used in Package Website.

• Add the wrapper functions for different common-used R objects in the spatial transcriptomics, such as spatialExperiment, etc; see the functions spe2seurat, spe2seuratList,seu2seuList and their help file in our developed SRTtools R package. This R package is designed to provide auxiliary functions for the serirs of our developed methods, such as SC-MEB, DR-SC and PRECAST.