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
	* Dot plots
	* Bar plots
	* Histograms
	* How to save plots
	* Volcano plots

## Getting started

* 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*

* Add a layer (with the **+** sign) that describes what kind of plot you want.

## 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}
# 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}
ggplot(data= df1 , mapping=aes(x=sample1, y=sample2)) + 
  geom_point() +
	ggtitle(label="my first ggplot") +
	geom_vline(xintercept=0)
```

Bookmark that [ggplot2 reference](https://ggplot2.tidyverse.org/reference/) and that good [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}
# Save in an object
p <- ggplot(data= df1 , mapping=aes(x=sample1, y=sample2)) +
	geom_point()
# 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}
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}
# Build a data frame from df1: add a column with "yes" and "no"
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()
```

## Box plots

* Simple boxplot showing the data distribution of sample 1:

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

* Split the data into 2 boxes:
```{r}
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 into a **long** format*

Plotting both sample1 and sample2:

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

```{r}
# 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()***

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

```{r}
# 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()**
  * **scale_fill_manual**

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

## Dot plots

Example of the expression of a gene in 6 samples: 2 experimental groups in triplicates.

```{r}
# create a named vector with the expression of a gene
mygene <- c(8.1, 8.2, 8.6, 8.7, 9.4, 8.5)

# the names of each element of the vector are sample names
names(mygene) <- c("KO1", "KO2", "KO3", "WT1", "WT2", "WT3")

# transform to long format
mygenelong <- melt(data=mygene)

# add new columns containing sample names and experimental groups
mygenelong$sample_name <- rownames(mygenelong)
mygenelong$group <- gsub("[1-3]{1}", "", mygenelong$sample_name)

# dot plot
  # add labels with "label" in the aes() and layer geom_text()
  # nudge_x adjust the labels horizontally
pdot <- ggplot(data=mygenelong, mapping=aes(x=group, y=value, col=group, label=sample_name)) + 
  geom_point() +
  geom_text(nudge_x=0.2)
```

* Add more layers:
  * **xlab()** to change the x axis label
  * **ylab()** to change the y axis label
  * **theme** to manage the legend
  
```{r}
pdot +  xlab(label="Experimental group") +
  ylab(label="Normalized expression (log2)") +
  ggtitle(label="Expression of gene 1") +
  theme_bw() +
  theme(legend.position="none")
  
```

## Bar plots

```{r}
# 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}
# Save the plot in the object "p"
pbar <- ggplot(data=df2, mapping=aes(x=grouping, fill=grouping)) + 
  geom_bar()

# 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"))

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

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

# Show intermediary and final plots
pbar
p2
p3
p4
```

## Histograms

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

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

```

Histogram on more samples (using df_long):

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

```

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

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

## About themes

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

```{r}
# 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}
png("myggplot.png")
p
dev.off()
```

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

```{r, eval=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}
# load package
library(gridExtra)
# 2 rows and 2 columns
grid.arrange(pscat, pbox, pbar, phist, nrow=2, ncol=2)
```

```{r, fig.width=10, fig.height=4,}
# 1 row and 4 columns
grid.arrange(pscat, pbox, pbar, phist, nrow=1, ncol=4)
```

**WARNING !!**: ggsave and grid.arrange are not directly compatible. To save a file organized by grid.arrange, use the regular functions (pdf, png etc.)

```{r, eval=FALSE}
jpeg("grid_arrange_plots.jpg")
grid.arrange(pscat, pbox, pbar, phist, nrow=1, ncol=4)
dev.off()
```