intro-to-gglot2.Rmd

---
title: "Introduction to ggplot2"
subtitle: "download @ bit.ly/2ZOShd4"
author: "Victor Yuan"
date: "2020-07-09"
output:
  xaringan::moon_reader:
    css: ["default", "middlebury-fonts", "custom1.css"]
    lib_dir: libs
    nature:
      ratio: 16:9
      highlightStyle: github
      highlightLines: true
      countIncrementalSlides: false
editor_options: 
  chunk_output_type: console
---

```{r setup, include = FALSE}
hook_source <- knitr::knit_hooks$get('source')
knitr::knit_hooks$set(source = function(x, options) {
  x <- stringr::str_replace(x, "^[[:blank:]]?([^*].+?)[[:blank:]]*#<<[[:blank:]]*$", "*\\1")
  hook_source(x, options)
})

knitr::opts_chunk$set(out.width="100%", out.height="100%")


ggplot2::theme_set(ggplot2::theme_gray(base_size=16))
```

# Set up

Install these packages

```{r eval = FALSE}
install.packages(tidyverse)
```

Load libraries

```{r}
library(tidyverse)
```

---

# Load gene expression / methylation data

```{r, message = FALSE}
geo_data <- read_csv('https://raw.githubusercontent.com/wvictor14/TOG/master/data/GSE98224.csv')

geo_data
```

---
layout: false
class: inverse center middle text-white

# 3 essential components 
## to every ggplot2 graph

### **Data**, **Geom**etry, **Aes**thetics

---

First step of every ggplot2 call is to *declare* the data.

.pull-left[
```{r our-first-plot-1, eval=FALSE}
ggplot(data = geo_data) #<<
```
]

.pull-right[
```{r our-first-plot-1-out, ref.label="our-first-plot-1", echo=FALSE}
```
]

---

Then, we can assign variables in our data to different *aesthetics* of the plot.

.pull-left[

```{r our-first-plot-2, eval=FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    #<<
           y = cg20970886)) #<<
```

This is referred to as *aesthetic mapping*.

]

.pull-right[
```{r our-first-plot-2-out, ref.label="our-first-plot-2", echo=FALSE}
```
]

---

Add **geometries (geoms)** to complete the plot.

.pull-left[


```{r our-first-plot-3, eval=FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) +
  geom_point() #<<
```

Geoms are like saying what type of plot you want (e.g. scatterplot, boxplots, histograms... etc.)

]

.pull-right[
```{r our-first-plot-3-out, ref.label="our-first-plot-3", echo=FALSE}
```
]

---

There are many *geoms*. Sometimes it makes sense to combine several.

.pull-left[
```{r our-first-plot-4, eval=FALSE, message = FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) +
  geom_point() + #<<
  geom_smooth(method = "lm") #<<
```
]

.pull-right[
```{r our-first-plot-4-out, ref.label="our-first-plot-4", echo=FALSE, message=FALSE}
```
]

---

We can assign other variables to other aesthetics, e.g. color.

.pull-left[
```{r our-first-plot-5, eval=FALSE, message = FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886,
           color = maternal_ethnicity)) + #<<
  geom_point() + 
  geom_smooth(method = "lm") 
```

But note that this assigned maternal ethnicity to the color of both points and lines!

]

.pull-right[
```{r our-first-plot-5-out, ref.label="our-first-plot-5", echo=FALSE, message=FALSE, warning = FALSE}
```
]

---

To assign color exclusively to points (and not lines), put inside specific geom:

.pull-left[
```{r our-first-plot-6, eval=FALSE, message = FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) + 
  geom_point(aes(color = maternal_ethnicity)) + #<<
  geom_smooth(method = "lm") 
```
]

.pull-right[
```{r our-first-plot-6-out, ref.label="our-first-plot-6", echo=FALSE, message=FALSE,warning=FALSE}
```
]

---

Can change the *shape* of points

.pull-left[
```{r our-first-plot-6-2, eval=FALSE, message = FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) + 
  geom_point(aes(color = maternal_ethnicity),
             shape = 3) + #<<
  geom_smooth(method = "lm") 
```

See [reference](https://ggplot2.tidyverse.org/reference/scale_shape.html) for complete list of shapes.

]

.pull-right[
```{r our-first-plot-6-2-out, ref.label="our-first-plot-6-2", echo=FALSE, message=FALSE,warning=FALSE}
```
]

---

A common mistake is to forget the aesthetic call.

.pull-left[
```{r our-first-plot-7, eval=FALSE, message = FALSE}
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) + 
  geom_point(color = "blue", #<<
             shape = 3) + 
  geom_smooth(method = "lm") 
```
]

.pull-right[
```{r our-first-plot-7-out, ref.label="our-first-plot-7", echo=FALSE, message=FALSE, warning = FALSE}
```
]

This assigns color to all the data

---

At this point, we've covered the 3 essential components to any ggplot2 plot:

1. **Data** - declare with a `ggplot(data = ...)` call 

2. **Aesthetics** - assign input to plot components with `aes()`, e.g. (x/y position, color)

3. **Geoms** - declare the type of geometry, e.g. `+ geom_point()` for points

---

# There are so many geoms

Each geom has their own required aesthetics, and optional ones

- `geom_point` requires `x` and `y`, and that they be numeric variables
- `geom_boxplot` requires `x` and `y`, but `x` must be categorical
- `geom_histogram` and `geom_density` requires `x`
- `geom_text` requires `x`, `y`, and `text`

Check out [tidyverse site](https://ggplot2.tidyverse.org/reference/#section-geoms) for full list.

You can visit help pages for more information on a specific geom's options (e.g. `?geom_point`)

Now we know the basics, we can explore ways to customize our plots

---
.left-code[

```{r fine-tune-1-1, eval=FALSE, message = FALSE}
ggplot(data = geo_data) #<<
```

We'll start by looking at the methylation of this CpG site between preeclamptic and non-preeclamptic samples

First we declare the data.

]

.right-plot[
```{r fine-tune-1-1-out, ref.label="fine-tune-1-1", echo=FALSE, message=FALSE, warning = FALSE, out.height="100%"}
```
]

PE: diagnosed with preeclampsia
---

.left-code[

```{r fine-tune-1-2, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis, #<<
           y = cg20970886, #<<
           fill = diagnosis))  #<<
```

Then we declare the mappings of our variables to aesthetics

]

.right-plot[
```{r fine-tune-1-2-out, ref.label="fine-tune-1-2", echo=FALSE, message=FALSE, warning = FALSE, out.height="100%"}
```
]

PE: diagnosed with preeclampsia
---

.left-code[

```{r fine-tune-1-3, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis, 
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() #<<
```

To specify we want boxplots, we use `geom_boxplot`

]

.right-plot[
```{r fine-tune-1-3-out, ref.label="fine-tune-1-3", echo=FALSE, message=FALSE, warning = FALSE, out.height="100%"}
```
]

PE: diagnosed with preeclampsia
---

.left-code[

```{r fine-tune-2-1, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_point() #<<
```

It can be informative to plot all individual data points over top of the boxplots.

To add individual data points, we simply add another geometry, `geom_point`

But it's a bit hard to see when the points overlap each other..

]

.right-plot[
```{r fine-tune-2-1-out, ref.label="fine-tune-2-1", echo=FALSE, message=FALSE, warning = FALSE}
```
]

---

.left-code[

```{r fine-tune-2-2, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() #<<
```

`geom_jitter` adds "noise" so that the points are spread out horizontally.

]

.right-plot[
```{r fine-tune-2-2-out, ref.label="fine-tune-2-2", echo=FALSE, message=FALSE, warning = FALSE}
```
]

---
layout: false
class: inverse center middle text-white

# Customizing your graphs
# Scales and themes

---

# Scales

`aes` determines which data variables are mapped to each component of the graph

`scale_*_*` functions determine *how* this mapping is done

`scale_<aes>_<type>` calls all start with "`scale_`" followed by the target aesthetic (e.g. x, y, color, fill), and finished by the type (e.g. discrete, continuous).

For example,

Want to change the limits on the y-axis? where the ticks appear? or maybe change to a log scale? Use 
`scale_y_continuous(limits = c(0,1))` or
`scale_y_log10()`

Want to change colors? Use
`scale_color_discrete()` for categorical variables
`scale_color_continuous()` for continuous variables

---

.left-code[

```{r fine-tune-3, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2"))  #<<
```

Here I assign specific colors to the categories of the diagnosis variable.

I supplied a vector of colors (can be in hex code) of same length of the number of categories of the variable `diagnosis`.

]


.right-plot[
```{r fine-tune-3-out, ref.label="fine-tune-3", echo=FALSE, message=FALSE, warning = FALSE}
```
]

---

.left-code[

```{r fine-tune-4-1, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("Controls", #<<
                              "Cases")) #<<
```

Here I change the labels of my x-axis.

]

.right-plot[
```{r fine-tune-4-1-out, ref.label="fine-tune-4-1", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---

.left-code[

```{r fine-tune-4-2, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", #<<
                              "PE" = "Cases")) #<<
```

It's better to be explicit about which label corresponds to which category

]

.right-plot[
```{r fine-tune-4-2-out, ref.label="fine-tune-4-2", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---


.left-code[

```{r fine-tune-5, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) +
  scale_y_continuous(limits = c(0, 1), #<< 
                     breaks = c(0, 0.5, 1))  #<<
```

Here I expand the y axis to 0 and 1, the natural range of methylation.

I also change where I want the ticks (i.e. "breaks") to appear.

Note that the y axis is a numeric variable and x axis is categorical, and how the respective scale calls reflect that.

]

.right-plot[
```{r fine-tune-5-out, ref.label="fine-tune-5", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---

.left-code[

```{r fine-tune-6, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) + 
  scale_y_continuous(limits = c(0, 1), 
                     breaks = c(0, 0.5, 1))  +
  theme(axis.text = element_text(colour = 'blue')) #<<
```

The **`theme()`** function call allows for a customization of the non-data components of a plot. Things like the title, labels, font size, gridlines, etc.

Pull up `?theme` to see a full description of all options

]

.right-plot[
```{r fine-tune-6-out, ref.label="fine-tune-6", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---

.left-code[

```{r fine-tune-7, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) +
  scale_y_continuous(limits = c(0, 1), 
                     breaks = c(0, 0.5, 1))  +
  theme(axis.text = element_text(colour = 'blue'), #<<
        panel.grid.major= element_line(colour = 'black'), #<<
        panel.grid.minor = element_blank()) #<<
```

Most `theme()` arguments will require an "`element_*`" as input.

The type of element depends on the type of input (e.g. `element_text` for `axis.text`, `element_rect` for `panel.border`).

`element_blank` to remove components.

]

.right-plot[
```{r fine-tune-7-out, ref.label="fine-tune-7", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---

.left-code[

```{r fine-tune-8, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) +
  scale_y_continuous(limits = c(0, 1), 
                     breaks = c(0, 0.5, 1))  +
  theme_bw(base_size = 20) #<<
```

There are some predefined themes that look nice and easy to use.

- `theme_gray` - default ggplot2 theme
- `theme_classic` - minimal with no gridlines
- `theme_bw` - clean look with white background

[List of complete ggplot2 themes](https://ggplot2.tidyverse.org/reference/ggtheme.html)

]

.right-plot[
```{r fine-tune-8-out, ref.label="fine-tune-8", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---

.left-code[

```{r fine-tune-9, eval=FALSE, message = FALSE}
ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) +
  scale_y_continuous(limits = c(0, 1), 
                     breaks = c(0, 0.5, 1))  +
  theme_bw(base_size = 20) +
  theme(legend.position = 'top') #<<
```

You can customize these complete themes by calling `theme()` after e.g. `theme_bw()`

]

.right-plot[
```{r fine-tune-9-out, ref.label="fine-tune-9", echo=FALSE, message=FALSE, warning=FALSE, out.width="100%"}
```
]

---

.left-code[

```{r save-1, eval=FALSE, message = FALSE}
p <- ggplot(data = geo_data, #<<
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) +
  scale_y_continuous(limits = c(0, 1), 
                     breaks = c(0, 0.5, 1))  +
  theme_bw(base_size = 20) +
  theme(legend.position = 'top')
```


]

.right-plot[

There are a couple of options to save plots in R.

Probably the simplest way is to use `ggsave` from `ggplot2`.

First thing to do is to assign your plot into an object.

I assigned our plot to the object named `p`

]

---

.left-code[

```{r save-2, eval=FALSE, message = FALSE}
p <- ggplot(data = geo_data, 
       aes(x = diagnosis,
           y = cg20970886,
           fill = diagnosis)) +
  geom_boxplot() +
  geom_jitter() +
  scale_fill_manual(values = c("orange", "#7ED7F2")) + 
  scale_x_discrete(labels = c("non-PE" = "Controls", 
                              "PE" = "Cases")) +
  scale_y_continuous(limits = c(0, 1), 
                     breaks = c(0, 0.5, 1))  +
  theme_bw(base_size = 20) +
  theme(legend.position = 'top')

ggsave(plot = p, #<<
       filename = "this-plot.png", #<<
       device = 'png', #<<
       dpi = 72, #<<
       height = 5, #<<
       width = 7) #<<
```

]

.right-plot[

Then we can call `ggsave` on object `p`.

I would recommend specifying the following options:

- `filename`, the name and location where you want the plot to be saved
-  `device`, the type of image file (e.g. "pdf", "png", "tiff", etc...)
- `height`, `width` - determines the dimensions of your plot
- `dpi`, resolution

After you run the code, check your local directory for the png file.

]
---

# Resources

- Stack exchange for online help
- TOG study group / slack
- [Past TOG workshops](https://github.com/BCCHR-trainee-omics-group/StudyGroup)
- [ggplot2  extensions](https://exts.ggplot2.tidyverse.org/)
- [ggplot2 cheatsheet](https://www.rstudio.com/wp-content/uploads/2015/03/ggplot2-cheatsheet.pdf)
- [r 4 data science data visualization chapter](https://r4ds.had.co.nz/data-visualisation.html)
- [Eva Maerey's ggplot2 grammar guide](https://evamaerey.github.io/ggplot2_grammar_guide/about)
---