CellProjectedPhenotypes (CPP) is a method to calculate phenotype scores for single cells based on their distance to phenotype-associated cell transcriptional neighborhoods.
Before installing the package, ensure that the required dependencies are installed:
install.packages("remotes")
remotes::install_github("immunogenomics/lisi")
BiocManager::install("sparseMatrixStats")
CPP makes extensive use of Milo, a method to calculate differential abundance on KNN graphs from single-cell datasets. More information about Milo can be obtained from their github page and in the associated manuscript.
CPP uses the Local Inverse Simpson's Index (LISI) to remove poorly mixed cells. Details of this method can be gained from this manuscript and from the corresponding github page.
The latest development version of this package can be installed from Github using devtools:
remotes::install_github("KellisLab/CellProjectedPhenotypes")
Here we demonstrate basic CPP usage by calculating cell projected phenotype scores from a Seurat-derived RDS file with corresponding metadata. Both of these files are available in our data directory.
- Load essential libraries
library(Seurat)
library(SeuratObject)
library(miloR)
library(SingleCellExperiment)
- Set input variables, including the name of the input file, corresponding metadata file, varsToTest (metadata variables to test for neighborhood associations), varsToCorrect(metadata covariates to control for in association testing), and col_name (the column of seurat object corresponding to the donor identifier).
file <- 'Vasculature_cells.rds'
meta <- read.csv("Metadata.csv")
varsToTest <- c("amyloid","tangles")
varsToCorrect <- c("fixation_interval","msex","pmi","study")
col_name <- "projid"
- Load single-cell data and filter and subsample individuals with extreme cell counts.
raw_cells <- readRDS(file)
seurat_obj <- CreateSeuratObject(counts = raw_cells@assays$RNA@counts)
seurat_obj[[col_name]] <- raw_cells[[col_name]]
seurat_obj$cell_type <- raw_cells$cell_type_high_resolution
seurat_obj <- filterAndSubsampleByCellCount(seurat_obj)
- Make a PCA embedding from the Seurat object and save first 12 principal components for later
seurat_obj <- NormalizeData(seurat_obj, normalization.method = "LogNormalize", scale.factor = 10000)
seurat_obj <- FindVariableFeatures(seurat_obj, selection.method = "vst", nfeatures = 2000)
seurat_obj <- ScaleData(seurat_obj, features = VariableFeatures(object = seurat_obj))
seurat_obj <- RunPCA(seurat_obj, features = VariableFeatures(object = seurat_obj))
cellPCs <- seurat_obj@[email protected][,1:12]
seurat_obj <- removePoorlyMixedCells(seurat_obj, cellPCs, col_name, 3)
- Construct a KNN graph, neighborhoods, and neighborhood cell counts with Milo
sce <- as.SingleCellExperiment(seurat_obj)
milo_obj <- Milo(sce)
milo_obj <- buildGraph(milo_obj, k = 30, d =12) # 12 PCA dimensions, 30 nearest neighbors
milo_obj <- makeNhoods(milo_obj, prop = 0.10, k = 30, d=12, refined = TRUE) # prop is proportion of vertices to sample for neighborhoods.
milo_obj <- countCells(milo_obj, meta.data = data.frame(colData(milo_obj)), samples=col_name)
- calculate neighborhood phenotype associations, the PCA coordinates of neighborhood centers, a matrix of distances between cells and neighborhoods, and a data frame of cell phenotype scores.
neighborhoodPhenotypeAssociations <- runNeighborhoodAssociationTest(milo_obj, varsToTest, varsToCorrect, col_name)
NeighborhoodCenters <- calculateNeighborhoodCenterMatrix(milo_obj, cellPCs)
cellNeighborhoodDistanceMatrix <- calculateCellNeighborhoodDistanceMatrix(NeighborhoodCenters, cellPCs, milo_obj)
cellScores <- calculateCellScores(milo_obj, neighborhoodPhenotypeAssociations, cellNeighborhoodDistanceMatrix, col_name, varsToTest)
We welcome contributions and suggestions from the community. Please report ideas to our issues page.
This package is licensed under the GPL-3 License.
- Gerard Bouland Author and maintainer
- Riley J. Mangan Author and contributor
- Manolis Kellis Author and oversight