ggplot2.Rmd

# ggplot2 package

```{r global_options, include=FALSE}
knitr::opts_chunk$set(fig.width=5, fig.height=4,
                      echo=TRUE, warning=FALSE, message=FALSE)
```

* Graphing package inspired by the **G**rammar of **G**raphics work of Leland Wilkinson.

* A tool that enables to concisely describe the components of a graphic.

* Why ggplot2 ?
	+ Flexible
	+ Customizable
	+ Pretty !
	+ Well documented

* We will see:
	* Scatter plots
	* Box plots
	* Bar plots
	* Histograms
	* How to save plots
	* Volcano plots

## Getting started

Load package:

```{r, eval=TRUE, echo=TRUE, warning=F, message=F}
library(ggplot2)
```

* All ggplots start with a **base layer** created with the **ggplot()** function:

```{r, eval=F, echo=TRUE}
ggplot(data=dataframe, mapping=aes(x=column1, y=column2))
```

*The base layer is setting the grounds but NOT plotting anything:*

* You then add a layer (with the **+** sign) that describes what kind of plot you want:
	* geom_point()
	* geom_bar()
	* geom_histogram()
	* geom_boxplot()
	* ...

* And then you will add **one layer at a time** to add more features to your plot!

## Scatter plot

```{r, eval=F}
# Example of a scatter plot: add the geom_point() layer
ggplot(data=dataframe, mapping=aes(x=column1, y=column2)) + geom_point()
```

* Example of a simple scatter plot:

```{r, eval=TRUE}
# Create a data frame
df1 <- data.frame(sample1=rnorm(200), sample2=rnorm(200))

# Plot !
ggplot(data=df1 , mapping=aes(x=sample1, y=sample2)) + 
	geom_point()
```


* Add **layers** to that object to customize the plot:
	* **ggtitle** to add a title
	* **geom_vline** to add a vertical line
	* etc.
```{r, eval=TRUE}
ggplot(data= df1 , mapping=aes(x=sample1, y=sample2)) + 
  geom_point() +
	ggtitle(label="my first ggplot") +
	geom_vline(xintercept=0)
```

Bookmark this [ggplot2 reference](https://ggplot2.tidyverse.org/reference/) and the [cheatsheet](https://www.rstudio.com/wp-content/uploads/2016/11/ggplot2-cheatsheet-2.1.pdf) for some of the ggplot2 options.

* You can save the plot **in an object** at any time and add layers to that object:

```{r, eval=TRUE}
# Save in an object
p <- ggplot(data= df1 , mapping=aes(x=sample1, y=sample2)) +
	geom_point()

# Show p: write p in the console and ENTER
p

# Add layers to that object
p + ggtitle(label="my first ggplot")
```

* What is inside the **aes** (aesthetics)function ?
	* Anything that varies according to your data !
		* Columns with values to be plotted.
		* Columns with which you want to, for example, color the points.

Color all points in red (not depending on the data):

```{r, eval=TRUE}
ggplot(data=df1 , mapping=aes(x=sample1, y=sample2)) +
	geom_point(color="red") 
```


Color the points according to another column in the data frame:

```{r, eval=TRUE}
# Build a new data frame df2 from df1: 
  # add a column "grouping" containing "yes" and "no" values.
df2 <- data.frame(df1, 
                  grouping=rep(c("yes", "no"), c(80, 120)))

# Plot and add the color parameter in the aes():
pscat <- ggplot(data=df2, mapping=aes(x=sample1, y=sample2, color=grouping)) + 
  geom_point()
pscat
```

Note that the legend is automatically added to the plot!

**HANDS-ON**

We will now use the **rock** dataset from the `datasets` package. It contains the measurements on 48 rock samples from a petroleum reservoir.

* Create a scatter plot of **area** versus **peri** (perimeter).
* Color the points according to column **perm** of **rock**
* Create a horizontal line representing the **median perimeter**.

<details>
<summary>
*Answer*
</summary>

```{r}
# Create a scatter plot of **area** versus **peri** (perimeter).
ggplot(data=rock, mapping=aes(x=area, y=peri)) + geom_point()

# Color the points according to column **perm** of **rock**
ggplot(data=rock, mapping=aes(x=area, y=peri, color=perm)) + 
    geom_point()

# Create a horizontal line representing the **median perimeter**.
ggplot(data=rock, mapping=aes(x=area, y=peri, color=perm)) + 
    geom_point() + 
    geom_hline(yintercept=median(rock$peri))
```

</details>

## Box plots

* Simple boxplot showing the data distribution of sample 1:

```{r, eval=TRUE}
ggplot(data=df2, mapping=aes(x="", y=sample1)) + geom_boxplot()
```

* Split the data into 2 boxes, depending on the **grouping** column:
```{r, eval=TRUE}
ggplot(data=df2, mapping=aes(x=grouping, y=sample1)) + geom_boxplot()
```

* What if you want to plot both sample1 and sample2 ?<br>
*You need to convert the data from a **wide** into a **long** format*
<br>

What is the **long** format ?<br>
One row **per observation/value**.

<img src="images/plots/wide2long.png" width="800">


Plotting both sample1 and sample2:

```{r, eval=FALSE}
# install package reshape2
install.packages("reshape2")
```

```{r, eval=TRUE, message=FALSE, warning=FALSE, error=FALSE}
# load package
library("reshape2")

# convert to long format
df_long <- melt(data=df2)
	# all numeric values are organized into only one column: value
# plot:
ggplot(data=df_long, mapping=aes(x=variable, y=value)) + 
  geom_boxplot()
```

* What if now you also want to see the distribution of "yes" and "no" in both sample1 and sample2 ?<br>
*Integrate a parameter to the **aes()***: either *color* or *fill*.

```{r, eval=TRUE}
# Either color (color of the box border)
ggplot(data=df_long, mapping=aes(x=variable, y=value, color=grouping)) + 
  geom_boxplot()
```

```{r, eval=TRUE}
# Or fill (color inside the box)
ggplot(data=df_long, mapping=aes(x=variable, y=value, fill=grouping)) + 
  geom_boxplot()
```

Do you want to change the default colors?<br>

* Integrate either layer:
	* **scale_color_manual()** for the boxes border color
	* **scale_fill_manual()** for the boxes color (inside)

```{r, eval=TRUE}
pbox_fill <- ggplot(data=df_long, mapping=aes(x=variable, y=value, fill=grouping)) + 
	geom_boxplot() +
	scale_fill_manual(values=c("slateblue2", "chocolate"))

pbox_fill

pbox_col <- ggplot(data=df_long, mapping=aes(x=variable, y=value, color=grouping)) +
        geom_boxplot() +
        scale_color_manual(values=c("slateblue2", "chocolate"))
pbox_col
```

**HANDS-ON**

Let's use the **CO2** dataset that represents the carbon dioxide uptake in grass plants:

* Create a boxplot that represents the **uptake** for each **Treatment**.
* Split each boxplot per **Type**. *Use either the color or the fill argument*.
* Move the legend to the bottom of the plot. You can get help from [this page](http://www.sthda.com/english/wiki/ggplot2-legend-easy-steps-to-change-the-position-and-the-appearance-of-a-graph-legend-in-r-software)

<details>
<summary>
*Answer*
</summary>

```{r}
# Create a boxplot that represents the **uptake** for each **Treatment**.
ggplot(data=CO2, mapping=aes(x=Treatment, y=uptake)) + geom_boxplot()
# Split each boxplot per **Type**.
ggplot(data=CO2, mapping=aes(x=Treatment, y=uptake, fill=Type)) + 
    geom_boxplot()
# Move the legend to the bottom of the plot.
ggplot(data=CO2, mapping=aes(x=Treatment, y=uptake, fill=Type)) + 
    geom_boxplot() +
    theme(legend.position = "bottom")
```

</details>

## Bar plots

```{r, eval=TRUE}
# A simple bar plot
ggplot(data=df2, mapping=aes(x=grouping)) + geom_bar()
```

* Customize:
  * **scale_x_discrete** is used to handle x-axis title and labels
  * **coord_flip** swaps the x and y axis
  
```{r, eval=TRUE}
# Save the plot in the object "p"
pbar <- ggplot(data=df2, mapping=aes(x=grouping, fill=grouping)) + 
  geom_bar()

pbar

# Change x axis label with scale_x_discrete and change order of the bars:
p2 <- pbar + scale_x_discrete(name="counts of yes / no", limits=c("yes", "no"))

p2

# Swapping x and y axis with coord_flip():
p3 <- p2 + coord_flip()

p3

# Change fill
p4 <- p3 + scale_fill_manual(values=c("yellow", "cyan"))

p4
```

**HANDS-ON**

Let's use the **chickwts** dataset again:

* Create a barplot of the different **feed supplements**.
* Change the orientation of the x-axis labels (Look it up in [this post](https://rstudio-pubs-static.s3.amazonaws.com/32217_6e2c396729704050972fb84f0d58ee22.html) ).

<details>
<summary>
*Answer*
</summary>

```{r}
# Create a barplot of the different **feed supplements**.
ggplot(data=chickwts, mapping=aes(x=feed)) + 
  geom_bar()

# Change the orientation of the x-axis labels
ggplot(data=chickwts, mapping=aes(x=feed)) + 
  geom_bar() +
  theme(axis.text.x=element_text(angle=45))
```

</details>
  
  
### Bar plots with error bars

We can create error bars on barplots.
<br>
Let's create a toy data set, that contains 7 independent qPCR measurements for 3 genes:

```{r, eval=TRUE}
pcr <- data.frame(Dkk1=c(18.2, 18.1, 17.8, 17.85, 18.6, 12.4, 10.7),
                  Pten=c(15.1,15.2, 15.0, 15.6, 15.3, 14.8, 15.9),
                  Tp53=c(9.1, 9.9, 9.25, 8.7, 8.8, 9.3, 7.8))
```

The height of the bar will represent the **average** qPCR measurement. The error bar will represent the **average - standard deviation** on the lowe part, and the **average + standard deviation** on the high part.
<br>
We need to create a data frame that **summarizes** these values:

```{r, eval=TRUE}
pcr_summary <- data.frame(average=apply(pcr, 2, mean), 
                          standard_deviation=apply(pcr, 2, sd),
                          genes=colnames(pcr))
```

And now we can plot!

```{r, eval=TRUE}
ggplot(pcr_summary, aes(x=genes, y=average, fill=genes)) + 
  geom_bar(stat = "identity") +
  geom_errorbar(aes(ymin=average-standard_deviation, ymax=average+standard_deviation), colour="black", width=.1) 
```
        
## Histograms

Simple histogram on one sample (using the df2 data frame):

```{r, eval=TRUE}
ggplot(data=df1, mapping=aes(x=sample1)) + geom_histogram()
```

Histogram on more samples (using df_long):

```{r, eval=TRUE}
ggplot(data=df_long, mapping=aes(x=value)) + geom_histogram()
```

Split the data per sample ("variable" column that represents here the samples):

```{r, eval=TRUE}
ggplot(data=df_long, mapping=aes(x=value, fill=variable)) + geom_histogram()
```

By default, the histograms are **stacked**: change to position **dodge** (side by side):

```{r, eval=TRUE}
phist <- ggplot(data=df_long, mapping=aes(x=value, fill=variable)) + 
	geom_histogram(position='dodge')
phist
```

**HANDS-ON**

Going back to the **rock** dataset:

* Create a histogram of the rocks **perimeter**.
* Add a density plot to the histogram, following instructions from [this post](http://www.sthda.com/english/wiki/ggplot2-histogram-plot-quick-start-guide-r-software-and-data-visualization#add-mean-line-and-density-plot-on-the-histogram)

<details>
<summary>
*Answer*
</summary>

```{r}
# Create a histogram of the rocks **perimeter**.
ggplot(data=rock, mapping=aes(x=peri)) + geom_histogram()
# Add a density plot to the histogram
ggplot(data=rock, mapping=aes(x=peri)) + 
    geom_histogram(aes(y=..density..)) +
    geom_density(alpha=.2, fill="lightblue") 
```

</details>

## About themes

You can change the default global **theme** (background color, grid lines etc. all non-data display):

```{r, eval=TRUE, message=F, warning=F}
# go back to a previous plot
p <- ggplot(data=df_long, mapping=aes(x=value)) + geom_histogram()

# Try different themes
p + theme_bw()
p + theme_minimal()
p + theme_void()
p + theme_grey()
p + theme_dark()
p + theme_light() 
```

## Saving plots in files

* The same as for regular plots applies:

```{r, eval=F, message=F, warning=F}
png("myggplot.png")
p
dev.off()
```

* You can also use the ggplot2 **ggsave** function:

```{r, eval=F, message=F, warning=F}
# By default, save the last plot that was produced
ggsave(filename="lastplot.png")

# You can pick which plot you want to save:
ggsave(filename="myplot.png", plot=p)

# Many different formats are available: 
  # "eps", "ps", "tex", "pdf", "jpeg", "tiff", "png", "bmp", "svg", "wmf"
ggsave(filename="myplot.ps", plot=p, device="ps")

# Change the height and width (and their unit):
ggsave(filename="myplot.pdf", 
  width = 20, 
  height = 20, 
  units = "cm")
```

* You can also organize several plots on one page
	* One way is to use the **gridExtra** package:
	* ncol, nrow: arrange plots in such number of columns and rows
```{r, eval=F}
install.packages("gridExtra")
```

```{r, eval=TRUE, message=F, error=F, warning=F}
# load package
library(gridExtra)
# 2 rows and 2 columns
grid.arrange(pscat, pbox_fill, pbar, phist, nrow=2, ncol=2)
```

```{r, fig.width=10, fig.height=4, eval=TRUE, message=F, warning=F}
# 1 row and 4 columns
grid.arrange(pscat, pbox_fill, pbar, phist, nrow=1, ncol=4)
```

Combine ggsave and grid.arrange:

```{r, eval=FALSE}
myplots <- grid.arrange(pscat, pbox_fill, pbar, phist, nrow=1, ncol=4)

ggsave(filename="mygridarrange.png", plot=myplots)
```

**HANDS-ON** (same as for the "base" plots!)

Go back to the previous plots you created (if you didn't save commands in an R script, you can refer to the *History* tab in the top-right panel):

* Save 1 plot of your choice in a **jpeg** file.
* Save 3 plots of your choice in a **pdf** file (one plot per page).
* Organize the same 3 plots in 1 row / 3 columns. Save the image in a **png** file. Play with the **width** (and perhaps also **height**) argument of `png()` until you are satisfied with the way the plot renders.

<details>
<summary>
*Answer*
</summary>

```{r}
# Save 1 plot of your choice in a **jpeg** file.
phisto <- ggplot(data=df_long, mapping=aes(x=value)) + geom_histogram()
ggsave(filename="mygridarrange.jpeg", plot=phisto)
# Save 3 plots of your choice in a **pdf** file (one plot per page).
pbarplot <- ggplot(data=chickwts, mapping=aes(x=feed)) + geom_bar()
pboxplot <- ggplot(data=CO2, mapping=aes(x=Treatment, y=uptake, fill=Type)) + geom_boxplot()
pdf("my3ggplot2.pdf")
phisto
pbarplot
pboxplot
dev.off()
# Organize the same 3 plots in 1 row / 3 columns.
my3ggplots <- grid.arrange(phisto, pbarplot, pboxplot, nrow=1, ncol=3)
ggsave(filename="my3ggplots.png", plot=my3ggplots, width=12)
```

</details>