scAI: a single cell Aggregation and Integration method for analyzing single cell multi-omics data

scAI is an unsupervised approach for integrative analysis of gene expression and chromatin accessibility or DNA methylation proflies measured in the same individual cells.
scAI infers a set of biologically relevant factors, which enable various downstream analyses, including the identification of cell clusters, cluster-specific markers and regulatory relationships.
scAI provides an intuitive way to visualize features (i.e., genes and loci) alongside the cells in two dimensions.
scAI aggegrates chromatin profiles of similar cells in an unsupervised and iterative manner, which opens up new avenues for analyzing extremely sparse, binary scATAC-seq data.

Once the single cell multi-omics data are decomposed into multiple biologically relevant factors, the package provides functionality for further data exploration, analysis, and visualization. Users can:

Visualize the latent biological patterns of the multi-omics data
Visualize both genes and loci alongside cells onto the same two-dimensional space
Identify cell clusters from the inferred joint cell loading matrix and cluster-specific markers
Visualize clusters and gene expression in the low-dimensional space such as VscAI, t-SNE and UMAP
Infer regulatory relationships between cluster-specific chromatin region and marker genes

Overview of scAI

Check out our paper (Suoqin Jin#, Lihua Zhang# & Qing Nie*, Genome Biology, 2020) for the detailed methods and applications.

Packages

scAI has been implemented as both R package and MATLAB package under the license GPL-3. In each package, we provide example workflows that outline the key steps and unique features of scAI. The MATLAB package and examples are available here.

Installation of R package

Install from Github using devtools

devtools::install_github("sqjin/scAI")

Install from R source codes

Download source codes here and type (in R)

install.packages(path_to_file, type = 'source', rep = NULL) # The path_to_file would represent the full path and file name

This website shows other ways for building and installing an R package.

Examples and Walkthroughs

All the R markdown used to generate the walkthroughs can be found under the /examples directory.

Simulated single cell RNA-seq and ATAC-seq data (Walkthrough): This simulated data were generated based on bulk RNA-seq and DNase-seq profiles from the same sample using MOSim package.
Simulated single cell RNA-seq and ATAC-seq dataset 8 (Walkthrough): This simulated data set consists of five imbalanced cell clusters with five clusters in scRNA-seq data and three clusters in scATAC-seq data.
Paired single cell RNA-seq and ATAC-seq data of A549 cells (Walkthrough): This data describes lung adenocarcinoma-derived A549 cells after 0, 1 and 3 hours of 100 nM dexamethasone treatment.
Paired single-cell RNA-seq and single-cell methylation data of mESC (Walkthrough): This data describes the differentiation of mouse embryonic stem cells (mESC).
Paired single cell RNA-seq and ATAC-seq data of Kidney cells (Walkthrough): This data describes various subpopulations of Kidney cells, including scRNA-seq and scATAC-seq data of 8837 co-assayed cells.

Suggestions for speeding up on large-scale datasets

Using the Python implementation of scAI model

object <- run_scAI(object, K, do.fast = TRUE)

Feature selection

Feature selection can reduce the running time in both scAI model and downstream analysis such as dimension reduction.

Using informative genes for scRNA-seq data:

The most informative genes can be selected based on their average expression and Fano factor (see our paper for details).

object <- selectFeatures(object, assay = "RNA")
object <- run_scAI(object, K, do.fast = TRUE, hvg.use1 = TRUE)

Using informative loci for scATAC-seq or single cell methylation data:

Unlike scRNA-seq data, the largely binary nature of scATAC-seq data makes it challenging to perform ‘variable’ feature selection. One option is to select the nearby chromsome regions of the informative genes.

object <- selectFeatures(object, assay = "RNA")
loci.use <- searchGeneRegions(genes = object@var.features[[1]], species = "mouse")
object@var.features[[2]] <- loci.use
object <- run_scAI(object, K, do.fast = TRUE, hvg.use1 = TRUE, hvg.use2 = TRUE)

Another option is to use only the top n% of features or remove features present in less that n cells. This method is used in Signac.

Additional installation steps (possibly)

Please consider install RcppEigen and rfunctions if they are not automatically installed.
```
if(!require(devtools)){ install.packages("devtools")}
install.packages("RcppEigen")
devtools::install_github("jaredhuling/rfunctions")
```
Troubleshooting: Installing RcppEigen and rfunctions on R>=3.5 requires Clang >= 6 and gfortran-6.1. For MacOS, it's recommended to follow guidance on the official R page here OR the post. For Windows, please ensure that Rtools is installed.
Install other dependencies

scAI provides functionality for further data exploration, analysis, and visualization. A couple of excellent packages need to be installed.

library(devtools)
install_github('linxihui/NNLM')
install_github("yanwu2014/swne")
install_github("jokergoo/ComplexHeatmap")

Install Leiden algorithm for identifying cell clusters: pip install leidenalg. Please check here if there is any trouble.
Install UMAP and FIt-SNE for faster dimension reduction in reducedDims

Using UMAP and FIt-SNE is recommended for computational efficiency when using reducedDims on very large datasets.

-- install UMAP Python package: pip install umap-learn. Please check here if there is any trouble.

-- install FIt-SNE R package: Installing and compiling the necessary software requires the use of FIt-SNE and FFTW. For detailed instructions of installation, please visit this page.

Troubleshooting on the R Compiler Tools for Rcpp on macOS

If you get the error "clang: error: unsupported option '-fopenmp'" when installing R package, please consider the configuration in ~/.R/Makevars and see this post for detailed configuration. In addition, you may can also reinstall your R because -fopenmp option is usually added by R automatically if openmp is available.

If you are using macOS Mojave Version (10.14) and you might get the error "/usr/local/clang6/bin/../include/c++/v1/math.h:301:15: fatal error: 'math.h' file not found", please check the post. This error can be solved if running the following on the terminal:

sudo installer -pkg \
/Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg \
-target /

Help

If you have any problems, comments or suggestions, please contact us at Suoqin Jin (suoqin.jin@uci.edu) or Lihua Zhang (lihuaz1@uci.edu).

How should I cite scAI?

Jin, S., Zhang, L. & Nie, Q. scAI: an unsupervised approach for the integrative analysis of parallel single-cell transcriptomic and epigenomic profiles. Genome Biol 21, 25 (2020). https://doi.org/10.1186/s13059-020-1932-8

sqjin / scAI

readme