Installing New Packages

Materials adapted from Adrien Osakwe, Larisa M. Soto and Xiaoqi Xie.

1. Installing New Packages

Think of R as a smartphone. It comes with “factory settings” (Base R), but to do specific things like high-quality plotting or RNA-seq analysis, you need to download “Apps” (Packages).

1.1 CRAN (The “App Store”)

CRAN is the official repository for R. It is highly regulated and stable.

install.packages(c("dplyr", "ggplot2", "gapminder", "medicaldata"))
Note

  • dplyr: The “Swiss Army Knife” for data cleaning and table manipulation.

  • ggplot2: The gold standard for creating publication-quality figures (Nature/Science style).

  • gapminder: A famous dataset about global life expectancy and GDP.

  • medicaldata: A collection of real-world clinical trial datasets.

1.2 BioConductor (The Bioinformatics Specialized Store)

BioConductor is specifically for biological data (Genomics, Proteomics, etc.). It has its own release cycle and installation manager.

# 1. Install the Manager first
if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

# 2. Use the Manager to install tools like DESeq2
BiocManager::install("DESeq2")
Note

DESeq2: The standard tool for analyzing Differential Gene Expression from RNA-seq data.

NoteWhat does that if sentence mean?

This is a “Smart Installation” script. It checks if you already have the package before trying to download it again.

  1. require("BiocManager"): This tries to load the package. If it succeeds, it returns TRUE. If the package is missing, it returns FALSE.

  2. The ! (NOT) symbol: This flips the result.

    • If you HAVE the package (TRUE), it becomes FALSE → The if block is skipped.

    • If you LACK the package (FALSE), it becomes TRUE → The install.packages line runs.

  3. Result: It ensures your code doesn’t waste 10 minutes re-installing a package you already have every time you run the script.

1.3 GitHub (The Developer’s Lab)

If a package is very new or in development, it might only be on GitHub. You need the devtools or remotes package to “clone” it.

# devtools is not a default part of R; you must install it first
if (!require("devtools", quietly = TRUE))
    install.packages("devtools")

devtools::install_github("YuLab-SMU/ggmsa")
Note

ggmsa: A specialized tool for visualizing Multiple Sequence Alignments (MSAs) using ggplot2 logic.

1.4 Choosing Your Source

  • CRAN: Use for general data science (cleaning data, standard plots).

  • BioConductor: Use for any “omics” data. This is where you find tools for Bulk/Single-cell RNA-seq.

  • GitHub: Use only if the package isn’t available elsewhere or if you need the absolute latest (but potentially “buggy”) version.

1.5 Finding the Right Tool

  • Bioconductor Views: A great place to browse by topic (e.g., “Epigenetics”).

  • RDocumentation.org: Search for any function name to see which package it belongs to.

  • Google/Stack Overflow: Usually, if you search “How to do X in R,” the top result will suggest a package.

1.6 Important Notes on Installation

When installing, R might ask: “Do you want to update other packages?”

  • Recommendation: Usually, choose “None” or “No” if you are in a hurry, as updating everything can take a long time and occasionally break other code.

  • The “Library” Step: Remember, install.packages() puts the app on your phone, but library() is what actually opens the app so you can use it.

1.7 The “Double Colon” Trick

Normally, we use library(package) to load all the tools from a package. However, you can access a specific function directly using the package::function() syntax.

Why use this?

  1. Preventing Naming Conflicts: Both the stats package (built into R) and the dplyr package have a function called filter(). If you load both, R will “mask” the older one. Using dplyr::filter() ensures R knows exactly which “filter” tool you want to use.

  2. Clarity for Collaborators: When someone else reads your code, they might not know which package lfcShrink() belongs to. Writing DESeq2::lfcShrink() makes your code self-documenting.

  3. One-Time Use: If you only need a package once in your entire script (like BiocManager::install()), it is cleaner to use the double colon than to load the entire library into your computer’s memory.

# Example 1: Installing without loading the manager
BiocManager::install("DESeq2")

# Example 2: Being explicit about which 'filter' to use
# This is very common in bioinformatics pipelines!
clean_data <- dplyr::filter(my_results, pvalue < 0.05)

This is very common in bioinformatics scripts because it prevents “naming conflicts” (when two different packages have a function with the same name).

1.8 Checking Package Versions

In bioinformatics, package versions change rapidly. A result you get with DESeq2 version 1.38 might be slightly different in version 1.42.

To ensure your research is reproducible, you should always check which version you are using before writing your paper.

# Check the version of a specific package
packageVersion("base")

# Or check everything currently loaded in your session
sessionInfo()
Note

Note: The sessionInfo() command is a “best practice” to run at the end of every analysis. It records your R version, your operating system, and every package version used.