Practical - beginner

Project - set-up

• Create a new project in a meaningful folder name on your computer such as R_workshop/day1-beginner

using the project manager utility, top-right of the rstudio window.

project menu

• Create a new folder data using bottom-right panel > Files tab > New Folder button

Files tab

• Create a new script to write and execute your R commands. top-left panel > Create icon > New Script entry.

create menu

Now, you have the 4 panels of the rstudio layout.

• Save the script with a relevant name practical-beginner.R

Reading data

Download this simple tab-separated file http://lsru.github.io/r_workshop/data/women.tsv

and save it inside the folder R_workshop/day1-beginner/data.

Remember, your current active rstudio project should be day1-beginner

load it: All paths are relative to the root which is the projects folder

library("readr")
df <- read_tsv("data/women.tsv", col_names = TRUE)
df

##    height weight
## 1      58    115
## 2      59    117
## 3      60    120
## 4      61    123
## 5      62    126
## 6      63    129
## 7      64    132
## 8      65    135
## 9      66    139
## 10     67    142
## 11     68    146
## 12     69    150
## 13     70    154
## 14     71    159
## 15     72    164

Thanks to readr the object df is already a tibble diff rstudio blog: tibble

Manipulate a data frame

We keep this section short, as we will focus on dplyr to perform tasks on data frames

Access to one column, display only the first elements

head(df$height)

## [1] 58 59 60 61 62 63

Using a similar syntax, apply:

• the function mean() to find the mean of women’ height.

Solution

mean(women$height)

## [1] 65

• the function var() to find the variance of women’ weight.

Solution

var(women$weight)

## [1] 240.2095

To compute her BMI (remember height are inches and weight US pounds) the formula is:

\[BMI = \frac{weight}{height^2} * 703\]

For the first individual (^2 for square):

(115 / 58^2) * 703

## [1] 24.0324

• Compute the BMI for all individuals, save it as bmi

Solution

bmi <- (women$weight / women$height^2) * 703

• Compute the mean and median of all BMI

Solution

mean(bmi)

## [1] 22.72443

median(bmi)

## [1] 22.46272

plotting

First load dplyr. This enables the use of the %>% pipe operator

library("dplyr")

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

Using df dataset:

• plot the heigh in function of the weight (geom_point())

Solution

library("ggplot2")
df %>%
  ggplot(aes(x = weight, y = height))+
  geom_point()

• use the previous scatterplot, but map the point’ size to the bmi

Solution

df %>%
  ggplot(aes(x = weight, y = height,
             size = bmi))+
  geom_point()

tidying and plotting

df has 2 columns, both contain values.

• use gather() from tidyr to get two columns
  • measure for either height or weight
  • value for actual measurements

Remember that gather takes by default all columns.

• store the result into df_melt

Solution

library("tidyr")
df_melt <- gather(df, measure, value)

• plot the distribution as boxplots of both measures

Solution

df_melt %>%
  ggplot(aes(x = measure, y = value))+
  geom_boxplot()

plot densities

adding a column to a data frame

Let’s add bmi as a third column to df.

df$bmi <- bmi
head(df)

## Source: local data frame [6 x 3]
## 
##   height weight      bmi
##    (int)  (int)    (dbl)
## 1     58    115 24.03240
## 2     59    117 23.62856
## 3     60    120 23.43333
## 4     61    123 23.23811
## 5     62    126 23.04318
## 6     63    129 22.84883

plot densities

• Gather (from tidyr) the 3 columns and plot all densities using different colours and set them translucent You will need to make a new df_melt data frame first.

Solution

df_melt <- gather(df, measure, value)
df_melt %>%
  ggplot(aes(x = value, fill = measure, colour = measure))+
  geom_density(alpha = 0.7)

The 3 distributions have very different ranges.

• Plot the same data but faceting it by measure (Use the appropriate free scale).

Solution

df_melt %>%
  ggplot(aes(x = value, fill = measure, colour = measure))+
  geom_density(alpha = 0.7)+
  facet_wrap(~ measure, scale = "free")

When faceting, the 3 distributions are drawn in distinct plots: mapping the colours to measure is useless.

• redo the plot using a lightblue colour for all. Be careful to NOT set the colour inside aes().

Solution

df_melt %>%
  ggplot(aes(x = value))+
  geom_density(fill = "blue", alpha = 0.7)+
  facet_wrap(~ measure, scale = "free")

Supplementary exercices

reading more complex file

Microarray data from the GEO dataset GSE35982.

• download this compressed file: GSE35982.tsv.gz in your data folder.
• read it using read_tsv() and store it into a data frame named gse. The file will be uncompressed seamlessly.

Solution

gse <- read_tsv("data/GSE35982.tsv.gz")

• Is the file tidy?

Solution

No, since all samples (starts by “GSM*“) are in different columns

• Gather the samples. Look at the gather help page to select columns based on characters.

Solution

library("tidyr")
gse_melt <- gse %>%
  gather(sample, value, starts_with("GSM"))

• plot the distributions as boxplots

Solution

library("ggplot2")
gse_melt %>%
  ggplot(aes(x = sample, y = value))+
  geom_boxplot()+
  coord_flip()+
  theme_bw()

• Any obvious issues? Check the file and find out what happened.

Hint

the locale setting in readr allows to specify the decimal mark used for float numbers

• Correct the mistake by reading again the file with the adjusted relevant option and store the data into a a new object.
• Replace the wrong column in gse by the correct one found in the data frame you just created.

Solution

gsefr <- read_tsv("data/GSE35982.tsv.gz", locale = locale(decimal_mark = ","))
gse$GSM878683 <- gsefr$GSM878683

• tidy the samples again.

Solution

gse_melt <- gse %>%
  gather(sample, value, starts_with("GSM"))

• plot the distributions as boxplots

Solution

gse_melt %>%
  ggplot(aes(x = sample, y = value))+
  geom_boxplot()+
  coord_flip()+
  theme_bw()

• do the data appear normalised?

Solution

Yes, perfectly normalised

unilur Rmarkdown template - E. Koncina