Introduction to ggplot2

class: center, middle, inverse, title-slide

# Introduction to ggplot2
## download @ bit.ly/2ZOShd4
### Victor Yuan
### 2020-07-09

---

# Set up

Install these packages

```r
install.packages(tidyverse)
```

Load libraries

```r
library(tidyverse)
```

```
## -- Attaching packages --------------------------------------- tidyverse 1.3.0 --
```

```
## v ggplot2 3.3.2     v purrr   0.3.4
## v tibble  3.0.5     v dplyr   1.0.3
## v tidyr   1.1.2     v stringr 1.4.0
## v readr   1.4.0     v forcats 0.5.0
```

```
## -- Conflicts ------------------------------------------ tidyverse_conflicts() --
## x dplyr::filter() masks stats::filter()
## x dplyr::lag()    masks stats::lag()
```

---

# Load gene expression / methylation data

```r
geo_data <- read_csv('https://raw.githubusercontent.com/wvictor14/TOG/master/data/GSE98224.csv')

geo_data
```

```
## # A tibble: 48 x 159
##    expr_geo_id meth_geo_id diagnosis tissue maternal_age maternal_bmi
##    <chr>       <chr>       <chr>     <chr>         <dbl>        <dbl>
##  1 GSM1940495  GSM2589532  PE        Place~           37         19.5
##  2 GSM1940496  GSM2589533  PE        Place~           40         25.7
##  3 GSM1940499  GSM2589534  PE        Place~           37         25  
##  4 GSM1940500  GSM2589535  PE        Place~           38         26.2
##  5 GSM1940501  GSM2589536  PE        Place~           33         31.2
##  6 GSM1940502  GSM2589537  PE        Place~           26         31.2
##  7 GSM1940505  GSM2589538  PE        Place~           31         18.6
##  8 GSM1940506  GSM2589539  PE        Place~           37         25.2
##  9 GSM1940507  GSM2589540  non-PE    Place~           35         18.6
## 10 GSM1940508  GSM2589541  PE        Place~           32         26.6
## # ... with 38 more rows, and 153 more variables: maternal_ethnicity <chr>,
## #   ga_weeks <dbl>, ga_days <dbl>, transcript_8033795 <dbl>,
## #   transcript_8103881 <dbl>, transcript_7904014 <dbl>,
## #   transcript_8127692 <dbl>, transcript_7990031 <dbl>,
## #   transcript_8121144 <dbl>, transcript_8150846 <dbl>,
## #   transcript_7962246 <dbl>, transcript_7941890 <dbl>,
## #   transcript_7896644 <dbl>, transcript_7992897 <dbl>,
## #   transcript_7973002 <dbl>, transcript_7979800 <dbl>,
## #   transcript_8112007 <dbl>, transcript_8036686 <dbl>,
## #   transcript_8001325 <dbl>, transcript_8180328 <dbl>,
## #   transcript_8109283 <dbl>, transcript_8041223 <dbl>,
## #   transcript_8144703 <dbl>, transcript_7997556 <dbl>,
## #   transcript_7955896 <dbl>, transcript_7939897 <dbl>,
## #   transcript_8035078 <dbl>, transcript_8113094 <dbl>,
## #   transcript_7893397 <dbl>, transcript_8110708 <dbl>,
## #   transcript_8102610 <dbl>, transcript_8083407 <dbl>,
## #   transcript_8174592 <dbl>, transcript_7922299 <dbl>,
## #   transcript_7979269 <dbl>, transcript_8074593 <dbl>,
## #   transcript_7967810 <dbl>, transcript_8052562 <dbl>,
## #   transcript_7927775 <dbl>, transcript_8005601 <dbl>,
## #   transcript_8129974 <dbl>, transcript_8070295 <dbl>,
## #   transcript_7952795 <dbl>, transcript_8044743 <dbl>,
## #   transcript_7896053 <dbl>, transcript_7894489 <dbl>,
## #   transcript_8048889 <dbl>, transcript_7894063 <dbl>,
## #   transcript_8171539 <dbl>, transcript_8011396 <dbl>,
## #   transcript_7983157 <dbl>, transcript_8171848 <dbl>,
## #   transcript_8097443 <dbl>, cg04950931 <dbl>, cg21697851 <dbl>,
## #   cg20092728 <dbl>, cg12804791 <dbl>, cg11619216 <dbl>, cg07802350 <dbl>,
## #   cg13175060 <dbl>, cg25632577 <dbl>, cg11811391 <dbl>, cg20981848 <dbl>,
## #   cg14025883 <dbl>, cg25493658 <dbl>, cg01491071 <dbl>, cg03777414 <dbl>,
## #   cg20586124 <dbl>, cg16175792 <dbl>, cg25961733 <dbl>, cg13912117 <dbl>,
## #   cg27307465 <dbl>, cg23825057 <dbl>, cg17949440 <dbl>, cg04098985 <dbl>,
## #   cg16886987 <dbl>, cg22860917 <dbl>, cg21594328 <dbl>, cg23903035 <dbl>,
## #   cg14393923 <dbl>, cg25103160 <dbl>, cg04640920 <dbl>, cg01522692 <dbl>,
## #   cg23249922 <dbl>, cg15903956 <dbl>, cg10688297 <dbl>, cg07989490 <dbl>,
## #   cg16090790 <dbl>, cg01519765 <dbl>, cg18444702 <dbl>, cg16404259 <dbl>,
## #   cg12077460 <dbl>, cg22517735 <dbl>, cg01713086 <dbl>, cg16734734 <dbl>,
## #   cg00886182 <dbl>, cg07891440 <dbl>, cg15715892 <dbl>, cg21368161 <dbl>,
## #   cg03766264 <dbl>, ...
```

---
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
*ggplot(data = geo_data)
```
]

.pull-right[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-1-out-1.png" width="100%" height="100%" />
]

---

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

.pull-left[

```r
ggplot(data = geo_data,
*      aes(x = ga_weeks,
*          y = cg20970886))
```

This is referred to as *aesthetic mapping*.

]

.pull-right[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-2-out-1.png" width="100%" height="100%" />
]

---

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

.pull-left[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-3-out-1.png" width="100%" height="100%" />
]

---

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

.pull-left[

```r
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) +
* geom_point() +
* geom_smooth(method = "lm")
```
]

.pull-right[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-4-out-1.png" width="100%" height="100%" />
]

---

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

.pull-left[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-5-out-1.png" width="100%" height="100%" />
]

---

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

.pull-left[

```r
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) + 
* geom_point(aes(color = maternal_ethnicity)) +
  geom_smooth(method = "lm") 
```
]

.pull-right[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-6-out-1.png" width="100%" height="100%" />
]

---

Can change the *shape* of points

.pull-left[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-6-2-out-1.png" width="100%" height="100%" />
]

---

A common mistake is to forget the aesthetic call.

.pull-left[

```r
ggplot(data = geo_data,
       aes(x = ga_weeks,    
           y = cg20970886)) + 
* geom_point(color = "blue",
             shape = 3) + 
  geom_smooth(method = "lm") 
```
]

.pull-right[
<img src="intro-to-ggplot2_files/figure-html/our-first-plot-7-out-1.png" width="100%" height="100%" />
]

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
*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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-1-1-out-1.png" width="100%" height="100%" />
]

PE: diagnosed with preeclampsia
---

.left-code[

```r
ggplot(data = geo_data, 
*      aes(x = diagnosis,
*          y = cg20970886,
*          fill = diagnosis))
```

Then we declare the mappings of our variables to aesthetics

]

.right-plot[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-1-2-out-1.png" width="100%" height="100%" />
]

PE: diagnosed with preeclampsia
---

.left-code[

```r
ggplot(data = geo_data, 
       aes(x = diagnosis, 
           y = cg20970886,
           fill = diagnosis)) +
* geom_boxplot()
```

To specify we want boxplots, we use `geom_boxplot`

]

.right-plot[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-1-3-out-1.png" width="100%" height="100%" />
]

PE: diagnosed with preeclampsia
---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-2-1-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-2-2-out-1.png" width="100%" height="100%" />
]

---
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
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-3-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-4-1-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-4-2-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-5-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-6-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-7-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-8-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
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[
<img src="intro-to-ggplot2_files/figure-html/fine-tune-9-out-1.png" width="100%" height="100%" />
]

---

.left-code[

```r
*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
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)
---