scCODE

scCODE (Consensus Optimization of Differentially Expressed gene detection for single cell)

Installation

Most dependent packages can be installed by the code below:

necessary1 <- c('doParallel', 'samr','doSNOW','pls')

installed <- necessary1 %in% installed.packages()[, 'Package']

if (length(necessary1[!installed]) >=1){

  install.packages(necessary1[!installed])

 }

necessary2<-c('DESeq2', 'DEsingle', 
          'edgeR', 'limma', 'MAST', 'S4Vectors', 'scDD', 'scmap', 'SingleCellExperiment', 'SummarizedExperiment')

installed <- necessary2 %in% installed.packages()[, 'Package']

if (length(necessary2[!installed]) >=1){

  if (!requireNamespace("BiocManager", quietly = TRUE))

    install.packages("BiocManager")

    library(BiocManager)

    BiocManager::install(necessary2[!installed])

  }

Two packages OGFSC and BPSC are required and need to install from Github.

First, check if we have already installed the two packages by :

'OGFSC'%in% installed.packages()

'BPSC'%in% installed.packages()

If not, we can download the installation file and install them through OGFSC and BPSC

Install scCODE

Now, we can install the scCODE by downloading the installation file in this page (scCODE_1.0.1.0.tar.gz), and install it:

install.packages("scCODE_1.2.0.0.tar.gz", repos = NULL, type="source")

Run scCODE

We can start using scCODE by running the sample data.

library(scCODE)

The input requires a count matrix (genes by cells), data_sccode contains the CD4+ T cells (naive and activated).

data1<-data1_sccode

data2<-data2_sccode

There are two parameters selectable. "light", True or False, default as True, run scCODE in a light version which saves time. "top_ranked", the number of top-ranked strategies selected (5-10), default as 5.

Now, we can run the sample data like below:

results<-scCODE(data1,data2,light = TRUE,top_ranked=5)

Output results

The results is a list containing the DE gene results and the metrics evaluation results.

The DE results is a dataframe which contains the ranked DE gene name along with its times-detected, logFC (group1/group2, and group 1 refers to the unique(group)[1]). it can be access by :

results$DE_results

The metrics results can be accessed by the code below, which is the Z-score of the metrics of each analysis strategies.

results$`Z-score-metrics`

Lotus plot, divided by detected_times

DE_results<-results$DE_results

times by detected times

times<- 1

DE_results$logFC_by_D<-ifelse(DE_results$logFC>=0,DE_results$logFC+(DE_results$Detected_times-1)*times,DE_results$logFC-(DE_results$Detected_times-  1)*times)
DE_results$Detected_times<-as.factor(DE_results$Detected_times)
library(ggplot2)
library(RColorBrewer)
coul <- rev(brewer.pal(5, 'Set1'))

P<-ggplot(
  DE_results, 
  aes(x = logFC_by_D, 
      y = -log10(P_adjust), 
      colour=Detected_times)) +
  geom_point(alpha=0.4, size=1.4) +
  scale_color_manual(values=coul)+
  labs(x="log2(fold change_by_detimes)",
       y="-log10 (p-value)")+
  theme_bw()+
  theme(plot.title = element_text(hjust = 0.5), 
        legend.position="right", 
        legend.title = element_blank()
  )

tiff('Fig lotus_sccode.tiff', units="in", width=8, height=8, res=300, compression = 'lzw')
P
dev.off()