This document provides an introduction to the statistical programming language R. It outlines what R is, how to access and use its interface, and how to work with basic data types like vectors, matrices, and factors. It also demonstrates how to import and export data, perform basic plotting and graphics, and gives examples working with biological data from Affymetrix chips. The presenter encourages attendees to ask questions and notes they are not a perfect teacher.

1. Introduction to R
Basic Teaching module
EMBL International PhD Program
13-10-2010
Sander Timmer & Myrto Kostadima

2. Overview
What is R
Quick overview datatypes, input/output and
plots
Some biological examples
I’m not a particular good teacher, so please
ask when you’re lost!

3. What is this R thing?
R is a powerful, general purpose language
and software environment for statistical
computing and graphics
Runs on Linux, OS X and for the unlucky few
also on Windows
R is open source and free!

4. Start your R interface

5. Variables
x <- 2
x <- x^2
x
[1] 4

6. Vectors
Many ways of generating a vector with a range of numbers:
x <- 1:10
assign(“x”, 1:10)
x <- c(1,2,3,4,5,6,7,8,9,10)
x <- seq(1,10, by=1)
x <- seq(length = 10, from=1,by=1)
x
[1] 1 2 3 4 5 6 7 8 9 10

7. Vectors
Common way to store multiple values
x <- c(1,2,4,5,10,12,15)
length(x)
mean(x)
summary(x)

8. Vectors
Vectors are indexed
x[5] + x[10]
[1] 15
x[-c(5,10)]
[1] 1 2 3 4 6 7 8 9

9. Matrices
Common form of storing 2 dimensional data
Think about having an Excel sheet
m = matrix(1:10,2,5)
[,1] [,2] [,3] [,4] [,5]
[1,] 1 3 5 7 9
[2,] 2 4 6 8 10
summary(m)

10. Factors
Factors are vectors with a discrete number of
levels:
x <- factor(c(“Cancer”, “Cancer”, “Normal”,
“Normal”))
levels(x)
[1] “Cancer” “Normal”
table(x)
Cancer Normal
2 2

11. Lists
A list can contain “anything”
Useful for storing several vectors
list(gene=”gene 1”, expression=c(5,2,3))
$gene
[1] “gene 1”
$expression
[1] 5, 2, 4

12. If-else statements
Essential for any programming language
if state then do x else do y
if(p < 0.01){
print(“Significant gene”)
}else{
print(“Insignificant gene”)
}

13. Repetition
You want to apply 1 function to every
element of a list
for(element in list){ ....do something.... }
For loops are easy though tend to be slow
Apply is the fast way of getting things done
in R:
apply(List,1,mean)

14. Data input
R has countless ways of importing data:
CSV
Excel
Flat text file

15. Data input
Most simple, the CSV file:
read.csv(“mydata.csv”,
row.names=T,col.names=T)
Load a tab separated file
read.table(“mytable.txt”, sep=”t”)
Load Rdata file
load(“mydata.Rdata”)

16. Data input
Also for more specific data sources:
Excel
Database connections
Mysql -> Ensembl e.g.
Affy
Affymetrix chips data
HapMap
.........

17. Data output
Most simple, the CSV file:
write.csv(x, file=”myx.csv”)
Save Rdata file:
save(x, file=”myx.Rdata”)
Save whole R session:
save(file=”mysession.Rdata”)

18. Graphics
Quick way to study your data is plotting it
The function “plot” in R can plot almost
anything out of the box (even if this doesn’t
make sense!)

19. plot(1:5,5:1)

20. plot(1:5,5:1, col=”red”, type=”l”)

21. plot(1:5,5:1, col=”red”, type=”l”,
main="Title of this plot",
xlab="x axis", ylab="y axis")

22. Basic graphics
With R you can plot almost any object
Multidimensional variables like matrixes
can be plotted with matplot()
Other often used plot functions are:
boxplot(), hist(), levelplot(), heatmap()

23. Advanced plotting

24. Advanced plotting

25. Advanced plotting

26. Before the example
Help page for functions in R can be called:
?plot, ?hist, ?vector
Examples for most functions can be runned:
example(plot)
Text search for functions can be done by
performing:
??plot

27. Example
Some example Affymetrix dataset to play
with
Checking distribution of data
Plotting data
Clustering data
Correlate data

28. Read file
library(affy)
library(affydata)
data(Dilution)
print(Dilution)

29. Read file
dil = pm(Dilution)[1:2000,]
dil.ex = exprs(Dilution)[1:2000,]
rownames(dil.ex) =
row.names(probes(Dilution))[1:2000]

30. Summary
Checking what we got
summary(dil)
mva.pairs(dil)
Or:
boxplot(log(dil.ex))
Or:
hist(dil.ex, xlim=c(0,500), breaks=1000)

31. We need to normalise
first
For almost all experiments you have to apply
some sort of normalisation
dil.norm = maffy.normalize(dil,
subset=1:nrow(dil))
colnames(dil.norm) = colnames(dil)
mva.pairs(dil.norm)

32. Most equal samples
Applying euclidian distance to detect most
equal samples
dil.norm.dist = dist(t(dil.norm))
dil.norm.dist.hc = hclust(dil.norm.dist)
plot(dil.norm.dist.hc)
Do the same for the non normalised dataset

33. Checking expression
Heatmap representation of expression levels
for different probes
heatmap(dil.ex.norm[1:50,])
You could apply a T-test for example to rank
to only plot the most significant probes

34. Checking expression
Heatmap representation of expression levels
for different probes
heatmap(dil.ex.norm[1:50,])
You could apply a T-test for example to rank
to only plot the most significant probes

35. Checking expression
You could apply a T-test for example to rank
to only plot the most significant probes
library(genefilter)
f = factor(c(1,1,2,2))
dil.exp.norm.t = rowttests(dil.exp.norm, fac=f)
heatmap(dil.exp.norm[order(dil.exp.norm.t
$dm)[1:10],])

36. Want to know more?
Using R will benefit all PhD’s in this room
Learning by doing
Loads of basic examples at:
http://addictedtor.free.fr/graphiques/
http://www.mayin.org/ajayshah/KB/R/
index.html
http://www.r-project.org/

37. Just keep in mind......

38. Questions?
Contact me:
swtimmer@ebi.ac.uk
http://www.ebi.ac.uk/~swtimmer/ for slides
or http://www.slideshare.net/swtimmer