Usefull functions
Operation on vector
Repeat a variable or vector X time
MyVector<-c(1,2,3,4,5) rep(MyVector,times=2)
## [1] 1 2 3 4 5 1 2 3 4 5
Or you can repeat each element of a vector X time
rep(MyVector,each=2)
## [1] 1 1 2 2 3 3 4 4 5 5
You can use paste to add some character (e.g. days)
paste("Day",MyVector,sep="_")
## [1] "Day_1" "Day_2" "Day_3" "Day_4" "Day_5"
Or collapse all character into a single character
paste("Day",MyVector,sep="_",collapse = ".")
## [1] "Day_1.Day_2.Day_3.Day_4.Day_5"
The apply functions
the apply functions can be usefull to use a function on every item of a vector, list, matrix
- sapply to use a function on item in a vector
MyGenes<-c("Arntl","Per2","Cry1","per1")
sapply(MyGenes,toupper)
## Arntl Per2 Cry1 per1 ## "ARNTL" "PER2" "CRY1" "PER1"
- lapply can be used on a list and will return list
lapply(MyGenes,toupper)
## [[1]] ## [1] "ARNTL" ## ## [[2]] ## [1] "PER2" ## ## [[3]] ## [1] "CRY1" ## ## [[4]] ## [1] "PER1"
- apply on a matrix
MyMatrix<-matrix(c(1,2,3,4,5,6),ncol=2) MyMatrix
## [,1] [,2] ## [1,] 1 4 ## [2,] 2 5 ## [3,] 3 6
# Compute mean value for each column apply(MyMatrix,2,mean)
## [1] 2 5
To compute for each row, use: apply(MyMatrix,1,mean)
For sum of matrix, you can simply use RowSum or ColSum as well as rowMeans and colMeans
rowSums(MyMatrix)
## [1] 5 7 9
rowMeans(MyMatrix)
## [1] 2.5 3.5 4.5
colSums(MyMatrix)
## [1] 6 15
colMeans(MyMatrix)
## [1] 2 5
aggregate function
You may want to compute some mean or sd over same samples or genes. You can use the aggregate function
MyData<-data.frame(X=1:4,
groups=rep(c("A","B"),each=2))
MyData
## X groups ## 1 1 A ## 2 2 A ## 3 3 B ## 4 4 B
aggregate(x = MyData$X, by = list(MyData$groups), FUN = mean)
## Group.1 x ## 1 A 1.5 ## 2 B 3.5
Create Functions
You can create your own function with some arguments
MyFunction<-function(arg1,arg2){
SumOfArgs<-arg1 + arg2
return(SumOfArgs)
}
MyFunction(arg1 = 1,arg2 = 10)
## [1] 11
You can have default arguments
MyOtherFunction<-function(arg1=3,arg2="Cortex"){
RepArg<-rep(arg2,arg1)
return(RepArg)
}
MyOtherFunction()
## [1] "Cortex" "Cortex" "Cortex"
You can pass any named or unamed argument using the ‘…’
MyPlotFunction<-function(x,...){
MyDatasquared<-x^2
plot(MyDatasquared,...)
}
The additional arguments are passed to the plot() function
MyPlotFunction(x = rnorm(10) , pch=19,
col = "red", main="Dots")
MyPlotFunction(x = rnorm(10) , type="l",
col="green", main="line")
Local and global variables
Within a function, your variable are local.
I can access within a function a variable defined outside a function (e.g. Variable_A)
Variable_A<-1
MyFunction<-function(){
Variable_B<-2
print(Variable_A)
}
MyFunction()
## [1] 1
however, Variable_B do not “exist” outside the function
exists("Variable_B")
## [1] FALSE
Also you cannot change Variable_A inside your function with a simple assignment operator ‘<-’
You need a superassignment operator ‘<<-’
Variable_A<-1
Variable_B<-1
MyFunction<-function(){
Variable_A<-2
Variable_B<<-2
return(invisible(NULL))
}
MyFunction()
Variable_A
## [1] 1
Variable_B
## [1] 2
Loops
You can use for loop
for (i in 1:10){
print(i)
}
## [1] 1 ## [1] 2 ## [1] 3 ## [1] 4 ## [1] 5 ## [1] 6 ## [1] 7 ## [1] 8 ## [1] 9 ## [1] 10
While loop
Num<-4
while(Num > 0){
Num<-Num - 1
print(Num)
}
## [1] 3 ## [1] 2 ## [1] 1 ## [1] 0
if else
Variable_A<- TRUE
x<-1
if (Variable_A==TRUE){
x<-x+1
y<-10
}else{
x<-x-1
}
x
## [1] 2
y
## [1] 10
gsub
If you want to replace some character into a string, use gsub (see regular expression)
gsub(pattern, replacement, x)
MyData<-c("Mus Musculus", "Homo Sapiens")
gsub(" ","_",MyData)
## [1] "Mus_Musculus" "Homo_Sapiens"
e.g. a more complex example
Pathways<-c("Glycolysis - Rattus norvegicus (rat)")
gsub(" - Rattus norvegicus \\(rat\\)","",Pathways)
## [1] "Glycolysis"
reshape2 package
Often, you will recieve data in the form of a data.frame or matrix, and you want to transform it into a data.frame with a single value per row.
A usefull package is reshape2 that will help you to do this:
MyDataFrame
## Genes Sample1 Sample2 Sample3 ## 1 Cry1 6 5 3 ## 2 Per2 3 5 9 ## 3 Arntl 2 8 5 ## 4 Cdk4 3 3 5
We can melt this
library(reshape2)
MyNewDF<-melt(MyDataFrame,id.vars = "Genes",
measure.vars = c("Sample1","Sample2","Sample3"),
variable.name = "Sample",value.name = "ReadCounts"
)
MyNewDF
## Genes Sample ReadCounts ## 1 Cry1 Sample1 6 ## 2 Per2 Sample1 3 ## 3 Arntl Sample1 2 ## 4 Cdk4 Sample1 3 ## 5 Cry1 Sample2 5 ## 6 Per2 Sample2 5 ## 7 Arntl Sample2 8 ## 8 Cdk4 Sample2 3 ## 9 Cry1 Sample3 3 ## 10 Per2 Sample3 9 ## 11 Arntl Sample3 5 ## 12 Cdk4 Sample3 5
Statistical test
Many statistical test are available in R:
student test: t.test()
non-parametric t-test: wilcox.test()
correlation: cor.test() with pearson,spearman or kendall methods
ANOVA: aov(), for type II or type III ANOVA you can use the the car package (car::Anova())
ANOVA post-hoc tukey using TukeyHSD()
Linear model using lm() (we will see how to write your model !)
Mixed effect model using lme4 package
Linear Regression
Linear model is written in this form:
\(y = \beta_1X_1 + \beta_2X_2 + \beta_3X_3 ...\)
where y is your variable of interest (dependent variable)
X your predictors (independent variables)
\(\beta\) values are the weights of your predictors (what you generally want to test)
plot(Age,Gene_Expression,pch=19)
We test for an effect of age on gene expression
summary(lm(Gene_Expression~Age))[[4]]
## Estimate Std. Error t value Pr(>|t|) ## (Intercept) 19.592825 4.4537217 4.399203 3.460734e-04 ## Age 2.016758 0.2054567 9.815975 1.187690e-08
And found a significant effect of age:
\(GeneExpression = \beta_0 + \beta_1 * Age + \epsilon\); \(\beta_1=\) 2.016758` 
Design Matrix
Depending on your design, DEseq2 may require a design matrix and which coefficient you want to test.
Your design may look like this:
## Gene_Expression Treatment Batch ## 1 0.16022185 Control B1 ## 2 -1.34087504 Control B2 ## 3 -0.50945013 Control B3 ## 4 -3.49290641 A_vehic B1 ## 5 1.49877604 A_vehic B2 ## 6 0.62524329 A_vehic B3 ## 7 0.03776714 Treatment B1 ## 8 0.30327340 Treatment B2 ## 9 -0.78182847 Treatment B3
You can write a model matrix that will look like this, but your Treatment A_vehic becomes the reference (your coefficient 4 represent the difference of \(Control - A\_vehic\), and coefficient 5: \(Treatment - A\_vehic\))
head(model.matrix(~Batch + Treatment,data=MyMetadata),9)
## (Intercept) BatchB2 BatchB3 TreatmentControl TreatmentTreatment ## 1 1 0 0 1 0 ## 2 1 1 0 1 0 ## 3 1 0 1 1 0 ## 4 1 0 0 0 0 ## 5 1 1 0 0 0 ## 6 1 0 1 0 0 ## 7 1 0 0 0 1 ## 8 1 1 0 0 1 ## 9 1 0 1 0 1
To solve this problem, write Treatment as factor and relevel it:
MyMetadata$Treatment<-factor(MyMetadata$Treatment,
levels = c("Control","A_vehic","Treatment"))
Now your coefficient 4 represent the difference of \(A\_vehic - Contro l\), and coefficient 5: \(Treatment - Control\))
head(model.matrix(~Batch + Treatment,data=MyMetadata),9)
## (Intercept) BatchB2 BatchB3 TreatmentA_vehic TreatmentTreatment ## 1 1 0 0 0 0 ## 2 1 1 0 0 0 ## 3 1 0 1 0 0 ## 4 1 0 0 1 0 ## 5 1 1 0 1 0 ## 6 1 0 1 1 0 ## 7 1 0 0 0 1 ## 8 1 1 0 0 1 ## 9 1 0 1 0 1
Files
Read files
You can read many different files in R (csv, txt, xlsx etc…)
If you file has a header, add header=T
You can use the built-in function like read.table() or read.csv() for text files
The package openxlsx will read excel format files
You can also use the Import Dataset in Rstudio (upper-right corner)
You can press tab while writing the path to a file to help you
You can also skip some line in a file using the skip option in read.table()
Write files
To write files use write.table() function.
Some option are very usefull, like:
write.table(MyDataFrame,
quote=F, # Avoid adding "" to character
sep="\t", # Use tab to separate data
col.names = T,# Add colnames to the file
row.names = T # Add rownames to the file
)
or save directly your variable in a Rdata object
save(MyDataFrame,MyDataFrame2,file="MyDataframes.Rdata")
and load the file using load()
load("MyDataframes.Rdata")