In phillipnicol/pcDiffPop: Robust Identification of Perturbed Cell Types in Single-cell RNA-seq

knitr::opts_chunk$set(echo = TRUE)

System Requirements

R is required to use scDist. In development, R version 4.0.0 and greater were used, but there may be compatibility with previous versions.

Installation

From the R console, devtools::install_github("phillipnicol/scDist"). Installation should take less than a minute on a standard machine.

Demo

The input to scDist is a normalized count matrix and correpsonding metadata that describes what condition and patient each cell belongs to. In this demo, we create a simulated dataset with 10 cell types. The demo should take less than a minute to run on a standard machine. The code is also modifiable to see how scDist performs for different parameter values.

library(scDist)
set.seed(1126490984)

Generate simulated data with 10 cell types and 5 patients in each group:

sim <- simData(nct=10,N1=5,N2=5)

dim(sim$Y) #Normalized counts
rownames(sim$Y) <- 1:1000
head(sim$meta.data)

Now we apply scDist:

out <- scDist(sim$Y,sim$meta.data,fixed.effects = "response",
              random.effects="patient",
              clusters="clusters")

The results data frame gives a summary of the estimated distance and uncertainty for each cell type

out$results

The true distances are

names(sim$D.true) <- letters[1:length(sim$D.true)]
sim$D.true

We can also plot the results

DistPlot(out)

To get a plot of genes that are associated with the perturbation use distGenes:

distGenes(out, cluster = "a")

Reference

If you use scDist in your work, please cite:

Nicol, P.B., Paulson, D., Qian, G., Liu, X.S., Irizarry, R.A., and Sahu, A.D. (2024). Robust identification of perturbed cell types in single-cell RNA-seq data. Nature Communications. Vol 15(7610). https://doi.org/10.1038/s41467-024-51649-3.

phillipnicol/pcDiffPop documentation built on Nov. 29, 2024, 6:06 p.m.