Approximate time: 30 min

Learning Objectives

Nested functions

Thus far, to perform any specific task, we have executed every function separately; if we wanted to use the results of a function for downstream purposes, we saved the results to a variable. As you become more comfortable with R, you will find that it is more efficient to code using nested functions, or functions within other functions, which will allow you to execute multiple commands at the same time.

Let’s start with an example from the previous lesson to demonstrate. To obtain the rows in metadata which correspond to celltype A, we used two lines of code:

idx <- metadata$celltype == "typeA"

metadata[idx, ]

However, we could have also done this in a single line of code and avoid having to create the variable idx:

metadata[metadata$celltype == "typeA", ]

This is a rather simple example combining only two lines of code, but you see how the code becomes lengthy but slightly more comprehensive. Even if you decide to avoid writing nested functions for the time being, you should still have experience reading and understanding them. The key to understanding nested functions is to read from the inside out.

Let’s work through some examples of nested functions!

Nested functions practice #1

Based on our metadata, how many samples are Wt and from celltype A?

Step 1: Classifying the samples using criteria based on logical operators:

vec <- metadata$celltype == "typeA" & metadata$genotype == "Wt"

Step 2: Identifying which samples meet our criteria (i.e the TRUE samples):

true_vec <- which(vec)

Step 3: Counting how many samples meet our criteria:

length(true_vec)

Nested code: Rather that assigning the output from each step to a separate variable, we could also just nest them into one another. To create a nested function, simply replace the variable name with the code on the right hand side of the assignment operator. Don’t forget to keep track of the sets of parentheses!

length(which(metadata$celltype == "typeA" & metadata$genotype == "Wt"))

Nested functions practice #2

You realize that you forgot to include important metadata regarding sex of your samples in your metadata file. You would like to add this data in and create a new metadata_new dataframe. Using functions separately would require us to execute three separate steps:

Step 1: Create the sex vector:

sex <- c("M","F","M","M","F","M","M","F","M","M","F","M")

Step 2: Turn the sex vector into a factor variable:

sex_fr <- factor(sex)

Step 3: Use the cbind function to add the column to the end of the metadata dataframe:

metadata_new <- cbind(metadata, gender=sex_fr)

Instead of performing all three steps, we would like to create a nested function. We could first replace the sex_fr variable with the assignment in Step2:

metadata_new <- cbind(metadata, gender=factor(sex))

Or we can go a step further and combine all steps, making your code slightly more difficult to read (we don’t recommend doing this):

metadata_new <- cbind(metadata,
			gender=factor(c("M","F","M","M","F","M","M","F","M","M","F","M")))

Nested functions practice #3

Now, let’s say that you are interested in finding out which samples (listed by sample name) in your dataset have “Wt” genotype within our metadata file. For our small dataset, we can simply do this by eye but for larger datasets it is easier to write code to do so:

Step 1: Determine the row locations in metadata for those samples with genotype equal to “Wt”:

rloc <- which(metadata$genotype == "Wt")

Step 2: Obtain the vector of row names from metadata:

rnames <- rownames(metadata)

Step 3: Identify the sample names by using the indices determined in Step 2:

wt_samples <- rnames[rloc]

Alternatively, we could combine the steps:

wt_samples <- rownames(metadata)[which(metadata$genotype == "Wt")]

Learning to understand nested functions is a critical part of your mastery of R. Not only will their use improve your efficiency, but nested functions are frequently encountered in help forums and R package documentation, so understanding them is critical to your learning process.

Exercise

Use nested functions and the subset() function to find out which samples (listed by sample name) in your dataset have “typeB” cell type within our metadata file.

This lesson has been developed by members of the teaching team at the Harvard Chan Bioinformatics Core (HBC). These are open access materials distributed under the terms of the Creative Commons Attribution license (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.