Promimity metrics in RΒΆ

This is an optional notebook that enables users to run proximity metrics that are only in R.

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("WGCNA")
BiocManager::install("reticulate",force=TRUE)
BiocManager::install("devtools",force=TRUE)
BiocManager::install("LoomExperiment",force=TRUE)
BiocManager::install("SingleCellExperiment",force=TRUE)
BiocManager::install("Seurat",force=TRUE, dependencies=TRUE)

library(devtools)
devtools::install_github("skinnider/dismay")
library(devtools)
devtools::install_github("cellgeni/sceasy",force = TRUE)

library('Seurat')
library('WGCNA')
library('dismay')

Options for proximity metrics in R are:

  • phi_s

  • rho_p

  • weighted_rank

  • zi-kendall

From the implementation provided by the dismay package (https://github.com/skinnider/dismay).

As dismay package computes metrics as similarities, they are converted to dissimilarities when merged with python metrics.

library(Seurat)
library(reticulate)
# Note here is where we suggest you use the conda env with scproximite installed
use_condaenv('scproximite')
library(sceasy)

data_dir = '../data/R_datasets/'

datasets = c("Discrete_Rare_Processed","Discrete_Abundant_Processed",
             "Continuous_Rare_Processed","Continuous_Abundant_Processed")

metrics = c('kendall') # 'phi_s', 'rho_p', 'weighted_rank','zi_kendall'

# Convert the datsets to rds format
for (data in datasets){
  name = paste(data_dir,data,sep="")
  sceasy::convertFormat(paste(name,".h5ad", sep=""), from="anndata", to="seurat", outFile=paste(name,".rds", sep=""))
}

Computes a distance matrix for each metric listed on each dataset provided, and outputs them to the data directory as .csv files for integration into the anndata object in python.

for (data in datasets){
  dat <- readRDS(paste(data_dir,data,".rds",sep =""))
  x <- as.matrix(dat@assays$RNA@data)
  for (m in metrics){
    out = dismay(x, m)
    nm = paste(data,m,'.csv', sep="_")
    write.csv(out, file=paste(data_dir, nm, sep=""))
  }
}