Week 13: Running R and Bioconductor in Google Colab

Google Colab has revolutionized the data science world, but it was originally designed, developed, and optimized for Python programmers. However, as bioinformatics and genomic data scientists, the vast majority of our favorite libraries and packages live in R and Bioconductor. Fortunately, with a few clever tricks, we can easily configure Google’s high-performance cloud servers to run R perfectly.

1. The Biological Problem

Imagine you have written a powerful R script that processes and analyzes bulk RNA-seq data. You want to run this pipeline on a massive dataset, but because of its scale, the files are stored directly in your personal Google Drive or within a shared team drive.

If you are using a standard cloud compiler, how do you get your code to “talk” to Google Drive? You do not want to download multiple gigabytes of genomic data onto your computer and then spend hours re-uploading them to your browser session every time you log in. To build seamless pipelines, we need our cloud-based R compiler to connect directly to our storage folders in Google Drive, reading raw matrices and writing output plots directly back to the cloud.

2. Intuition & Theory

To run R and Bioconductor smoothly in Google Colab, we have to overcome two distinct hurdles:

1. Swapping the Runtime (The R Kernel)

By default, whenever you open a new notebook in Google Colab, it loads a Python computational engine (known as a “kernel”). To write and compile R code, you have to tell Google Colab to swap its default engine with an active R kernel. This sets up an entirely new session pre-loaded with R, giving you access to all of R’s native statistical environments.

2. Handling the Ephemeral Cloud Environment

When you connect to Google Colab, Google spins up a temporary virtual computer just for you. This is an ephemeral environment, meaning that as soon as you close your browser tab or remain inactive for too long, Google reclaims that computer and erases its drive completely.

If you install a massive Bioconductor package, it will be wiped clean when the session ends. To prevent losing your analysis, we write code to mount Google Drive. Mounting acts exactly like plugging a virtual USB flash drive containing all your biological data directly into Google’s cloud server. This allows you to load files from Drive and securely save your output figures and matrices so they are never lost.

Source: [Wikimedia Commons/Jupyter Runtime Selection]

3. Visual Breakdown

To see how to initialize an R notebook, configure your Google Colab settings, and mount your Google Drive folders inside your code cells, watch this video:

4. Translating Theory to Code

Let’s look at the actual code structure used to run R in Colab. When you launch an R notebook, every code cell is compiled as pure R syntax.

Automatic Non-Interactive Installations

In interactive environments like RStudio, installing packages often prompts you to choose mirror sites or update dependencies. In a cloud script, these prompts will freeze your code cell indefinitely. We use the repos argument to make installations completely automatic and non-interactive:

# --- Installing Core R Packages in Colab ---

# Use a non-interactive installation with an explicit mirror repository
install.packages("ggplot2", repos = "https://cloud.r-project.org")

# Load the library to verify installation
library(ggplot2)

Mount and Access Google Drive

To read your biological sequences or clinical spreadsheets directly from Google Drive, we can use the R rgoogledrive package or standard system commands. Let’s look at how we typically load datasets directly from cloud resources or directories in a Colab notebook:

# --- Accessing Biological Data in the Cloud ---

# 1. Reading file directly from a public URL or server
gene_expression_url <- "https://raw.githubusercontent.com/bioconductor/bioc-data/main/expression_subset.csv"
expression_data <- read.csv(gene_expression_url)

# Display the first few rows of genomic expression
head(expression_data)

# 2. Saving your analysis directly to a custom file
# (If Google Drive is mounted at your instance, it can write straight to '/content/drive/My Drive')
write.csv(expression_data, file = "cleaned_expression_data.csv", row.names = FALSE)