Very quick introduction to R

In several exercises in this course, we use the statistical programming environment R. Here follows a very quick introduction, primarily for those who have never used R before. You can go through this now, or return to it when you first encounter R in the exercises.

First start R on your computer. How to do this depends on your operating system. If you are on a Linux or Mac OS X system, you would typically execute the command R at the shell prompt.

In this course, we are running R on a Linux server. If you would like to install R on your own computer, you can find the appropriate download for your system here.

When R starts, you should see a text message stating the version of R you are running and some further information, followed by a command-line “prompt” ( >, a greater-than sign). The prompt means that R is waiting for you to type a command. To get a feel for how this works, try out some arithmetic::

 > 2 + 3
 [1] 5
 > 5 * 4 + 10
 [1] 30

And some function calls

> abs(-5)
[1] 5
> sum(1,5,10)
[1] 16

To find out how to use a function, type its name preceded by a question mark::

> ?sin

This will bring up some help documentation for the function. You can use the arrow keys to scroll the help text up and down. Press q to get back to the R prompt.

Now try this:

> a <- 10

This command created an object called a. Objects are an important concept in R (as in many other programming languages), and we will be creating more of them in the RNA-seq exercises. We can inspect an object by just typing its name:

> a
[1] 10

We can also change the value of an object that we’ve created:

> a <- 2 * a	
> a
[1] 20

The object a created above is a vector with a single element. To create a vector with several elements, you can use the function c():

> b <- c(1, 2, 10)
> b
[1]  1  2 10

Or the colon operator::

> 1:10
[1]  1  2  3  4  5  6  7  8  9 10

A matrix can be created with the function cbind():

> b <- cbind(1:10, 101:110)
> b
   [,1] [,2]
[1,]    1  101
[2,]    2  102
[3,]    3  103
[4,]    4  104
[5,]    5  105
[6,]    6  106
[7,]    7  107
[8,]    8  108
[9,]    9  109
[10,]   10  110

We can then use indices to access selected elements of the matrix:

> b[1,]
[1]   1 101
> b[, 2]
[1] 101 102 103 104 105 106 107 108 109 110
> b[c(5,8), 2]
[1] 105 108

To end your R session, use the functions quit() or q():

> q()
Save workspace image? [y/n/c]: n

If you wish, you can save a workspace image. This means that the state of your R session is written to a file, which will be loaded the next time you start R in the same working directory. This is good if you want to continue the session at a later time. If not, just answer n (for no).

You can find manuals for R and more information on the R web site.

Working with R packages

R contains many functions, in particular for statistics and plotting. There are numerous additional functions and data sets, contained in R packages that are not installed by default.

Many such extension packages can be obtained from the Comprehensive R Archive Network (CRAN). We will also download packages from BioConductor, an alternative repository focused on R packages for bioinformatics.

Installing packages from CRAN

To install a CRAN package in R, use the install.packages() function. This simple command downloads the package from a specified repository (by default, CRAN) and installs it on your machine. For example, we can install a useful package called gplots:

> install.packages("gplots")

After installation, the package is stored in a location on your computer called the R package library. Packages are loaded from this library using the function library():

> library("gplots")

One useful function in the gplots package is heatmap.2(), which plots heatmaps with dendrograms. Since we have installed and loaded the gplots package, we can now use this function. Let’s try it with a matrix of random, normally distributed values:

> heatmap.2(matrix(rnorm(180),ncol=6))

Note that you only need to install a package once. The next time you start R, you can run library("gplots") directly, because the package has already been installed. If you upgrade to a new version of R, you may have to reinstall packages.

Installing packages from BioConductor

To install packages from BioConductor, you first have to load a function biocLite() from BioConductor and then use that function to install the packages. That is done in two lines of code:

> source("https://bioconductor.org/biocLite.R")
> biocLite("Name of package")

For example, let’s install the package DESeq2, which provides functions for RNA-seq differential expression analysis:

> source("https://bioconductor.org/biocLite.R")
> biocLite("DESeq2")

You should now be able to load the package DESeq2:

> library("DESeq2")

Installing packages from GitHub

Sometimes you want to install packages that are still under development. Those are mostly found on GitHub.

The example below shows how you can switch between a stable version and a development version of a package, in this case the package ggplot2. This example is provided for reference and you don’t have to understand it (or try it) for completing the RNA-seq labs.

First, we install and load the regular (stable) version of the ggplot2 package from CRAN:

> install.packages("ggplot2")
> library(ggplot2)

To see which packages are loaded, and their versions, use the function sessionInfo():

> sessionInfo()

Note the version number of ggplot2 reported by sessionInfo().

There is no entirely safe way to unload packages in R, so to work with a different package version, it is safest to quit and restart R. To quit:

> q()
Save workspace image? [y/n/c]: n

After starting R again, we install and load the devtools package, which provides functions for working with development versions of packages:

> install.packages("devtools")
> library(devtools)

Next we install the development version of ggplot2 from GitHub:

> dev_mode(on=TRUE)
> install_github("hadley/ggplot2")

Now we can load and work with the the development version of ggplot2:

> library(ggplot2)
> sessionInfo()

Check the output from sessionInfo() to verify that you now get a different version number for ggplot2!

When finished working with development versions, do:

> dev_mode(on=FALSE)

Note that if you only want to install from GitHub, you can ignore the dev_mode function calls above.