SCENA is a R package for unsupervised clustering of single cell RNA-Seq data
Version: 1.0.3
Depends: R(>3.6)
Import packages: parallel, doParallel,foreach, SNFtool, gpuR, apcluster, mclust, ggplot2
Citation: Consensus Clustering of Single-cell RNA-seq Data by Enhancing Network Affinity, Briefings in Bioinformatics 2021, DOI: 10.1093/bib/bbab236
requiredPackages = c('parallel','doParallel','foreach','ggplot2','SNFtool',"apcluster","mclust")
for(p in requiredPackages){
if(!require(p,character.only = TRUE)) install.packages(p)
library(p,character.only = TRUE)
}
install.packages("devtools")
devtools::install_github("shaoqiangzhang/SCENA")
install.packages("gpuR") ## see Note1
##Note1: the package "gpuR" was built on Linux x86_64 (https://www.rdocumentation.org/packages/gpuR), and cannot be installed on a Windows system.
To speed up matrix computing, recommend you install the "OpenBLAS" library on your operating system(Linux as an example)
git clone https://github.com/xianyi/OpenBLAS.git
cd OpenBLAS
make FC=gfortran
sudo make PREFIX=/usr/local install
then,set the number of threads using environment variables as follows
export OPENBLAS_NUM_THREADS=4
There are several detailed examples in the "examples" folder. You can directly run these examples in R.
The "data" folder includes the download URLs of some datasets
Step 1: load the package, download the dataset (rows are genes and columns are cells), and read file.
library(SCENA)
furl<-"https://s3.amazonaws.com/scrnaseq-public-datasets/manual-data/yan/nsmb.2660-S2.csv"
download.file(furl,destfile="./nsmb.2660-S2.csv")
Express=read.csv("./nsmb.2660-S2.csv", header = T,row.names = 1) ##see Note 2
Express=Express[,2:91] #select 90 cells. You can omit it if you use full data
##Note2: If the file suffix is ".txt" or ".tsv" , please use "read.table" instead of "read.csv".
Step 2: preprocess the input data as follows.
Express=datapreprocess(Express,log=T) #log=T is to do log-transformation, log=F is no log-transformation
At the end of the preprocessing step, suggestions for setting parameters are provided.
Step 3: do clustering in parallel with 5 CPU cores as follows.
detectCores()
cl <- makeCluster(5) # call 5 cpu cores
parLapply(cl,1:5,K=10,T=50,X1=50,X2=100,X3=150,X4=200,X5=250,Express=Express,select_features)##see Note3
stopCluster(cl)
##Note3: K is the number of K-nearest neighbors; T is the number of matrix iterations, X1~X5 are top number of selected features. For a big dataset(number of cells>1000), if you just need an acceptable result quickly, you can set the number of iterations T to be small (e.g. T=20).
Alternatively, if your computer supports GPU computing, you can do clustering in parallel with CPU+GPU as follows.
library(gpuR)
source('./parallelclust_gpu.R')## see Note 4
cl <- makeCluster(5)
parLapply(cl,1:5,K=10,T=50,X1=50,X2=100,X3=150,X4=200,X5=250,Express=Express,select_features_GPU)
stopCluster(cl)
##Note4: Because the GPU code cannot be called from the installed SCENA package directly, please copy it to your working path and run it using ’source'.
Step 4: do consensus clustering as follows.
cc=consClust(Express)
or
cc=consClust(Express,6) #set the number of clusters =6
For easy of use, Steps 3 and 4 for CPU computing have been integrated into a new function "scena_cpu" as follows.
cc=scena_cpu(Express) ##use default parameters
or
cc=scena_cpu(Express,T=20) ## "T" is the the number of matrix iterations.
or
cc=scena_cpu(Express,T=20,num=6) ##"num" is the number of clusters
plotPCA(Express,cc) # 'cc' is label of the predicted clusters
library(mclust)
presetlabel=substring(colnames(Express),1,4) ## read the column names as preset cell types
adjustedRandIndex(presetlabel,as.vector(cc)) ## 'cc' is predicted label
Authors: Shaoqiang Zhang, Yaxuan Cui, Yong Chen
Maintainer: Shaoqiang Zhang ([email protected])
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent