Advanced Data Visualization Homework

Part 1

We can start by taking a look at the distribution of the expression values.

1. Plot a histogram of the expression variable, and set the bin number equal to 100.

ggplot(ydat, aes(x = expression)) +
    geom_histogram(bins = 100) 

2. Check the distribution of each nutrient in the data set by adjusting the fill aesthetic. Use the same bin number for this histogram.

ggplot(ydat, aes(x = expression, fill = nutrient)) +
    geom_histogram(bins = 100)

Wow. That’s ugly. Might be a candidate for accidental aRt but not very helpful for our analysis.

3. Now split off the same histogram into a faceted display with 3 columns.

ggplot(ydat, aes(x = expression, fill = nutrient)) +
    geom_histogram(bins = 100) +
    facet_wrap(~ nutrient, ncol = 3)

The basic exploratory process above confirms that the overall distribution (as well each distribution by nutrient) is normal.

Part 2

Let’s compare the genes with the highest and lowest average expression values.

We can figure out which these are using some familiar logic:

1. Take the original ydat data frame …
2. Then group by symbol …
3. Then summarize mean of all expression values for that symbol …
4. Then arrange descending by the mean …
5. Then filter for the first or last row.

The code below implements that pipeline in dplyr syntax:

ydat %>%
    group_by(symbol) %>%
    summarise(meanexp = mean(expression)) %>%
    arrange(desc(meanexp)) %>%
    filter(row_number() == 1 | row_number() == n())

## # A tibble: 2 x 2
##   symbol meanexp
##    <chr>   <dbl>
## 1   HXT3    4.01
## 2   HXT6   -2.68

The output tells us that the gene with the highest mean expression is HXT3, while the gene with the lowest mean expression is HXT6.

4. Subset the data to only include these genes, and create a stripplot that has expression values as “jittered” points on the y-axis and the gene symbols the x-axis.

HINT you can add a “jitter” position to geom_point() but it’s easier to control width of the effect if you use geom_jitter()

gene_exp <- 
    ydat %>%
    filter(symbol == "HXT3" | symbol == "HXT6")

ggplot(gene_exp, aes(x = symbol, y = expression)) +
    geom_jitter(width = 0.1)

5. Now map each observation to its nutrient by color and adjust the size of the points to be 2.

ggplot(gene_exp, aes(x = symbol, y = expression)) +
    geom_jitter(aes(col = nutrient), width = 0.1, size = 2)

Although these two genes are on opposite ends of the distribution of average expression values, they both seem to express similar amounts when Glucose is the restricted nutrient.

Part 3

Now let’s try to make something that has a little bit more of a polished look.

6. Using dplyr logic, create a data frame that has the mean expression values for all combinations of rate and nutrient (hint: use group_by() and summarize()). Create a plot of this data with rate on the x-axis and mean expression on the y-axis and lines colored by nutrient.

nutrient_rates <-
    ydat %>%
    group_by(rate, nutrient) %>%
    summarise(meanexp = mean(expression)) 

ggplot(nutrient_rates, aes(x = rate, y =  meanexp)) + geom_line(aes(col=nutrient))

7. Add black dotted line (lty=3) that represents the smoothed mean of expression across all combinations of nutrients and rates.

ggplot(nutrient_rates, aes(x = rate, y =  meanexp)) + 
  geom_line(aes(col=nutrient), lty=1) + 
  geom_smooth(col = "black", lty=3, se = FALSE) 

8. Change the scale to include breaks for all of the rates.

HINT The read_csv() function read in the rate variable as continuous rather than discrete. There are a few ways to remedy this, but first see if you can set the scale for the x axis variable without changing the dataframe.

ggplot(nutrient_rates, aes(x = rate, y =  meanexp)) + 
  geom_line(aes(col=nutrient), lty=1) + 
  geom_smooth(col = "black", se = FALSE, lty=3) + 
  scale_x_continuous(breaks = nutrient_rates$rate)

9. By default ggplot() will name the x and y axes with names of their respective variables. You might want to apply more meaningful labels. Change the name of the x-axis to “Rate”, the name of the y-axis to “Mean Expression” and the plot title to “Mean Expression By Rate (Brauer)”

HINT ?labs will pull up the ggplot2 documentation on axes labels and plot titles.

ggplot(nutrient_rates, aes(x = rate, y =  meanexp)) + 
  geom_line(aes(col=nutrient), lty=1) + 
  geom_smooth(col = "black", se = FALSE, lty=3) + 
  scale_x_continuous(breaks = nutrient_rates$rate) + 
  xlab("Rate") + ylab("Mean Expression") + 
  ggtitle("Mean Expression By Rate (Brauer)")

10. Add a theme from the ggthemes package. The plot below is based on Edward Tufte’s book The Visual Display of Quantitative Information. Choose a theme that you like, but choose wisely – some of these themes will override other adjustments you’ve made to your plot above, including axis labels.

HINT 1: library(ggthemes) not working for you? Install the package first.

HINT 2 You can either do this by trial-and-error or check out the package vignette to get an idea of what each theme looks like: https://github.com/jrnold/ggthemes

library(ggthemes)
ggplot(nutrient_rates, aes(x = rate, y =  meanexp)) + 
  geom_line(aes(col=nutrient), lty=1) + 
  geom_smooth(col = "black", se = FALSE, lty=3) + 
  scale_x_continuous(breaks = nutrient_rates$rate) + 
  xlab("Rate") + ylab("Mean Expression") + 
  ggtitle("Mean Expression By Rate (Brauer)") + 
  theme_tufte()

11. The last step is to save the plot you’ve created. Write your plot to a 10 X 6 PDF using a ggplot2 function.

p <- ggplot(nutrient_rates, aes(x = rate, y =  meanexp)) + 
  geom_line(aes(col=nutrient), lty=1) + 
  geom_smooth(col = "black", se = FALSE, lty=3) + 
  scale_x_continuous(breaks = nutrient_rates$rate) + 
  ylab("Mean Expression") + 
  ggtitle("Mean Expression By Rate (Brauer)") + 
  theme_tufte()

ggsave(p, filename = "plot.pdf", width = 10, height = 6)