day5.Rmd

---
title: "Intro to R 2025 day 5"
author: "Matt Cannon"
date: '2025-04-04'
output:
  html_document:
    code_folding: hide
    toc: true
    toc_depth: 2
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

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

# Day 5

## General stuff

### geoms as layers
- When you're making a plot using ggplot2, think about it as if you're making a canvas and then adding layers to change it
- The layers are added in the order you add the code
- Layers can do things like add points, lines, or more abstract things like change the font of parts of the plot
- Each layer is separated by a "+" sign, which is pretty much the only time you'll see this in R
- The documentation for ggplot2 is quite good
    - https://ggplot2.tidyverse.org/reference/
    - One trick: skim the documentation for a plot that looks similar to what you want, then look at the code they used to make it
    - Another: You can also do a google image search for ggplot and the type of plot you want to make then go to the page where the image is from. Often times you'll find the code they used to make it.

### Anatomy of a ggplot

ggplot(                         # This makes your "canvas"
    data_frame_here,            # The data you're using - must be a long form data.frame or tibble
    aes(                        # "Aesthetics" - x, y, colors, etc  - Applies to all layers unless overwritten
        x = x here,             # The name of the column that has your x values - note: don't put this in quotes
        y = y here,             # The name of the column that has your y values
        other stuff             # You can specify a lot of things here - check the documentation
    )
) +
    layer_1 +                       # Layers can be things like geom_point(), geom_line(), geom_boxplot(), etc
    layer_2 +                       # You can add multiple different types of geoms, or even multiple of the same type
    layer_3 +
    other_stuff +                   # You can add other things like themes, labels, etc
    more_stuff +
    change_how_something_looks +
    modify_the_plot_somehow         
    

You can really customize the plot and the code can get quite long. Thinking about it as layers can help keep track of what you're doing



I'm not going over all the things in this document, just the ones marked with lots of pound signs ##############################################################

I put the rest in here for reference. If you go through it I've tried to drop in a bunch of random bits of code that might be useful to you in the future.

## Functions for today's activity, also for reference

#### geom_point() ##############################################################
Add a scatterplot layer to the plot
```{r}
ggplot(
  mtcars,
  aes(
    x = hp,
    y = carb
  )
) +
  geom_point()

# Set the overall look of the plot - I like the white background better than gray
# This will apply to all plots you make after you run this code, so you only need to run it once per session
theme_set(theme_bw()) # Changes default colors
theme_update(plot.title = element_text(hjust = 0.5)) # Force title to be in the center

ggplot(
  mtcars,
  aes(
    x = hp,
    y = carb
  )
) +
  geom_point() +
  theme_bw()
```

#### aes() ################################################
Change colors/shapes/etc

- fill     = color of the **inside** of a shape
- color    = color of the **border** of a shape, note that points use color for the color of the point
- shape    = the shape of the point
- size     = the size of the point or line
- alpha    = transparency of the point or line
- linetype = the type of line
- group    = the group that the data is in, used for things like boxplots
- label    = Text to plot somewhere, used in geom_text(), geom_label(), etc

If you put the argument outside of aes()

-   applies to everything in that layer
-   does not show up in the legend

If you put it inside aes()

-   Applies to parts of the layer
-   Shows up in legend

If it is in the aes() call inside of ggplot() `ggplot(aes(...))`

-   It applies to all layers
-   Unless another layer overwrites it

```{r}
mtcars %>%
  ggplot(
    aes(
      x = as.factor(cyl),
      y = hp
    )
  ) +
  geom_boxplot(
    color = "red", # These specifications aren't in aes()
    fill = "blue", # They apply to everything and we give specific values
    size = 4
  )

ggplot(
  mtcars,
  aes(
    x = hp, # These specifications are in aes()
    y = carb, # They apply to parts of the plot
    color = as.factor(vs), # They show up in the legend
    shape = as.factor(gear), # We give column names that have the groups we'll use
    size = disp
  )
) +
  geom_point()
```

#### geom_histogram()
- Better looking than hist()
- Specify how many bars to show with `bins`
```{r}
hist(storms$pressure, n = 200)

ggplot(
  storms,
  aes(x = pressure)
) +
  geom_histogram(bins = 200)
```

#### geom_density()
- Smooth your data out in a smoothed histogram-looking plot
    - The adjust argument can make it more or less smooth - 0 is the raw data, 1 is very smooth
- This can actually hide some of the nuance in your data
```{r}
ggplot(
  storms,
  aes(
    x = pressure,
    fill = as.factor(category)
  )
) +
  geom_density(alpha = 0.1, adjust = 0.5) + # alpha sets transparency -- 0 is clear, 1 is opaque
  scale_color_brewer(palette = "Set2") # The brewer color palettes are pretty, and "Set2" is color-blind friendly
```

#### geom_col() ################################################################
Add a barplot layer to the plot
```{r}
mtcars %>%
  rownames_to_column("car_make") %>% # This makes the row names a column so we can use it in the plot
  mutate(car_make = str_remove(car_make, " .+")) %>% # This removes everything after the first space in the car_make column
  ggplot(
    aes(
      x = car_make,
      y = hp
    )
  ) +
  geom_col(color = "red") +
  theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
  ggtitle("this is a title") +
  theme_update(plot.title = element_text(hjust = 0.5))

mtcars %>%
  rownames_to_column("car_make") %>%
  ggplot(aes(y = car_make, x = hp)) +
  geom_col()
```

#### geom_errorbar()###########################################################
Add error bars as a layer to plot
```{r}
summarized_hp <-
  mtcars %>%
  group_by(cyl) %>%
  summarize(
    mean_hp = mean(hp),
    sd_hp = sd(hp)
  )

summarized_hp

ggplot(
  summarized_hp,
  aes(
    x = as.factor(cyl),
    y = mean_hp
  )
) +
  geom_col() +
  geom_errorbar(
    aes(
      ymin = mean_hp - sd_hp,
      ymax = mean_hp + sd_hp
    ),
    width = 0.2,
    color = "red"
  )

# This has error bars in the back since that layer was added first!
ggplot(
  summarized_hp,
  aes(
    x = as.factor(cyl),
    y = mean_hp
  )
) +
  geom_errorbar(
    aes(
      ymin = mean_hp - sd_hp,
      ymax = mean_hp + sd_hp
    ),
    color = "red"
  ) +
  geom_bar(stat = "identity") +
  geom_point(color = "blue")
```

#### geom_jitter()
Drop-in replacement for geom_point()
Adds noise in x and y directions by default so you can see individual points
```{r}
ggplot(
  summarized_hp,
  aes(
    x = as.factor(cyl),
    y = mean_hp
  )
) +
  geom_col() +
  geom_errorbar(
    aes(
      ymin = mean_hp - sd_hp,
      ymax = mean_hp + sd_hp
    )
  ) +
  geom_jitter(
    data = mtcars, # Note that we can provide an individual geom it's own data
    aes(
      x = as.factor(cyl),
      y = hp
    ),
    color = "red", height = 0
  )
```

#### ggbeeswarm package
Adds noise in x direction and pulls things to the center to make it look nicer than jitter
```{r}
ggplot(
  summarized_hp,
  aes(
    x = as.factor(cyl),
    y = mean_hp
  )
) +
  geom_col() +
  geom_errorbar(
    aes(
      ymin = mean_hp - sd_hp,
      ymax = mean_hp + sd_hp
    )
  ) +
  ggbeeswarm::geom_beeswarm(
    data = mtcars,
    aes(
      x = as.factor(cyl),
      y = hp
    ),
    color = "red"
  )
```

#### geom_boxplot() ############################################################
Boxplots!
```{r}
mtcars %>%
  ggplot(aes(x = as.factor(cyl), y = hp)) +
  geom_boxplot() +
  geom_violin(adjust = 1, fill = "red", alpha = 0.5)
```

#### geom_smooth
Add a regression line to your plot
Can specify the type of regression with `method`
Can remove gray confidence interval with `se = FALSE`
```{r}
ggplot(
  mtcars,
  aes(
    x = mpg,
    y = disp
  )
) +
  geom_point() +
  geom_smooth()
```

#### Use multiple datasets in the same plot
Contents of the columns used as axes have to match
Each geom using it's own data needs to have "data = " in it
```{r}
ggplot(
  summarized_hp,
  aes(
    x = as.factor(cyl),
    y = mean_hp
  )
) +
  geom_col() +
  geom_errorbar(
    aes(
      ymin = mean_hp - sd_hp,
      ymax = mean_hp + sd_hp
    ),
    width = 0.5
  ) +
  geom_jitter(
    data = mtcars,
    aes(
      x = as.factor(cyl),
      y = hp
    ),
    color = "red"
  )
```

#### labs()
Add labels
```{r}
ggplot(mtcars) +
  geom_point(aes(x = hp, y = disp)) +
  labs(
    x = "X label goes here",
    y = "Y label goes here",
    title = "This is a title",
    subtitle = "Subtitle!",
    caption = "Cannon et al, 2034!",
    tag = "A"
  )
```

#### facet_wrap() ##############################################################
```{r}
dplyr::storms

ggplot(
  dplyr::storms,
  aes(x = pressure)
) +
  geom_histogram(bins = 100) +
  facet_wrap(~ category + status)

ggplot(
  dplyr::storms,
  aes(x = pressure)
) +
  geom_histogram(bins = 100) +
  facet_wrap(~category, scales = "free_y")

ggplot(
  dplyr::storms,
  aes(x = pressure)
) +
  geom_histogram(bins = 100) +
  facet_wrap(
    ~category,
    nrow = 1,
    scales = "free_y"
  )
```

#### theme()
A good guide to what theme elements are what:
https://henrywang.nl/ggplot2-theme-elements-demonstration/

You can change just about any aspect of how the plot looks with theme()
You'll have to look up how to change each specific bit
```{r}
ggplot(
  storms,
  aes(
    x = pressure,
    fill = category
  )
) +
  geom_density(alpha = 0.5) + # alpha sets transparency - 0 is clear, 1 is opaque
  scale_color_brewer(palette = "Set2") + # The brewer color palettes are pretty, and "Set2" is color-blind friendly
  theme(
    legend.position = c(0.2, 0.9),
    legend.direction = "horizontal",
    legend.key.width = unit(2, "cm"),
    legend.title = element_text(face = "bold"),
    strip.background = element_rect(color = "white", fill = "white"),
    strip.text.x = element_text(size = 30, face = "bold"),
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    panel.border = element_blank(),
    axis.line.x = element_line(color = "black"),
    axis.text.y = element_blank(),
    axis.ticks.y = element_blank()
  )
```

### ggsave()
```{r}
ggsave(
  "examplePlot.png",
  width = 10,
  height = 8
)

plot_in_variable <-
  ggplot(
    dplyr::storms,
    aes(x = pressure)
  ) +
  geom_histogram(bins = 100) +
  facet_wrap(~category)

# ggsave will get the plot size from how big your "plot" windows is in the right pane if you don't specify it!!!!!
ggsave(
  "anotherExamplePlot.png",
  plot = plot_in_variable
)
```

### heatmap() ##################################################################
If you like heatmaps also check out the pheatmap package
```{r}
heatmap(as.matrix(mtcars))

# Scaling the categories instead
heatmap(as.matrix(mtcars), scale = "column")
pheatmap::pheatmap(as.matrix(mtcars))
```

### PCA
```{r}
mtcars <- mtcars

pca_raw <- prcomp(mtcars, scale = TRUE)

?prcomp

# The output of prcomp is a list
# The element named "x" contains the actual principle components
# Though, it needs to be converted from a matrix to a dataframe for ggplot
pc_values <-
  pca_raw$x %>%
  as.data.frame() %>%
  rownames_to_column("car_make")

ggplot(
  pc_values,
  aes(
    x = PC1,
    y = PC2
  )
) +
  geom_point() +
  ggrepel::geom_text_repel(aes(label = car_make))
```

## Activity

## Recreate plots from a paper

- Data is from https://doi.org/10.1371/journal.pbio.2005756
    - Najafov A, Zervantonakis IK, Mookhtiar AK, Greninger P, March RJ, Egan RK, et al. (2018) BRAF and AXL oncogenes drive RIPK3 expression loss in cancer. PLoS Biol 16(8): e2005756. https://doi.org/10.1371/journal.pbio.2005756
- I downloaded their figure data and made it easier to import:
    - Fig1A_partial.txt
    - Fig1C.txt
    - Fig2F.txt
    - Fig3F.txt
- You can see what the plots should look like in the corresponding files in materials/Figure_***.png files
- You don't have to make them look exact, feel free to play around with how you want them to look

#### Figure 1A - geom_boxplot()
```{r}

```

#### Figure 1C - heatmap()
```{r}

```

#### Figure 2F - geom_point()
```{r}

```

#### Figure 3F - geom_col
use geom_errorbars()
and facet_wrap() to split the plots by cell_line
```{r}

```

## If you're super fast:
Keep going, do the rest 🙃

Download the data from the supplemental data here:

https://doi.org/10.1371/journal.pbio.2005756.s001


####
# You're done! You made it through introduction to R!
####

Suggestions for what to do next:
- Try to practice working in R as much as you can
- Find ways that R can be useful for you and explore packages that are useful for the type of work you do
- You're still learning! Don't get down on yourself when you struggle, because you will
    - The better you get at R, the less you'll struggle, but everyone runs into problems
- Read the documentation for functions/packages you think you'll use a lot
- Make all the errors!
    - Then google them and try to understand what went wrong
- Go back through and re-read the materials to reinforce what you've learned and pick up little details you may have missed
- If you're going to the SCRGOT single cell workshop, do some reading on Seurat
    - https://satijalab.org/seurat/
- The teams channel from the workshop shouldn't get deleted right away, so if you have problems, post there to see if people can help
    - Also, if other people post, read what problems they ran into and try to learn from it