scFeatures: Multi-view representations of single-cell and spatial data for disease outcome prediction

scFeatures is a tool that generates multi-view representations of single-cell and spatial data through the construction of a total of 17 feature types belonging to the following six categories.

1. cell type proportions
2. cell type specific gene expressions
3. cell type specific pathway expressions
4. cell type specific cell-cell interaction (CCI) scores
5. overall aggregated gene expressions
6. spatial metrics

Installation

The latest scFeatures can be installed using devtools:

library(devtools)
devtools::install_github("SydneyBioX/scFeatures")

The following dependency is needed for running the cell-cell communication feature type.

devtools::install_github("immunogenomics/presto")
devtools::install_github("jinworks/CellChat")

Quick start

scFeatures can be run using one line of code scfeatures_result <- scFeatures(data = data, sample = sample, celltype = celltype) which generates a list of dataframes containing all feature types in the form of samples x features.

Currently, scFeatures support scRNA-seq, spatial proteomics and spatial transcriptomics.

For scRNA-seq, run:

data("example_scrnaseq" , package = "scFeatures")
data <- example_scrnaseq
scfeatures_result <- scFeatures(data = data@assays$RNA@data, sample = data$sample, celltype = data$celltype,
type = "scrna", ncores = 8, species = "Homo sapiens")

For spatial proteomics, run:

# note, spatial data requires spatial coordinates of each cell. spatialCoords <- list( sample( 1:ncol(data), ncol(data)) , sample( 1:ncol(data), ncol(data) )) # generate fake coordinates
scfeatures_result <- scFeatures(data = data@assays$RNA@data, sample = data$sample, celltype = data$celltype,
type = "spatial_p", spatialCoords = spatialCoords, ncores = 8, species = "Homo sapiens")

For spatial transcriptomics, run:

# note, spatial data requires spatial coordinates of each cell. spatialCoords <- list( sample( 1:ncol(data), ncol(data)) , sample( 1:ncol(data), ncol(data) )) # as well as predicted probability of cell types in each spot spotProbability <- t(gtools::rdirichlet( ncol(data), rep(1, 5))) # simulate the cell type prediction result based on 5 cell types rownames( spotProbability) <- c("Cell type A", "Cell type B" , "Cell type C", "Cell type D", "Cell type E")
colnames( spotProbability ) <- colnames(data)
scfeatures_result <- scFeatures(data = data@assays$RNA@data, sample = data$sample, celltype = data$celltype,
type = "spatial_t", spatialCoords = spatialCoords, spotProbability = spotProbability, ncores = 8, species = "Homo sapiens")

Detailed vignette

Please see https://sydneybiox.github.io/scFeatures/articles/scFeatures_overview.html.

Reference

Cao, Y., Lin, Y., Patrick, E., Yang, P., & Yang, J. Y. H. (2022). scFeatures: multi-view representations of single-cell and spatial data for disease outcome prediction. In O. Vitek (Ed.), Bioinformatics (Vol. 38, Issue 20, pp. 4745–4753). Oxford University Press (OUP). https://doi.org/10.1093/bioinformatics/btac590