Abstract:

This video captures a Bioconductor Developer Forum dedicated to integrating the Rust programming language with R for bioinformatics applications, primarily using the extendr framework. The presentation is led by extendr maintainers Josiah and Masa, along with bioinformatics researcher Wuji, who shares a user's perspective.

The forum begins by introducing Rust as a high-performance, memory-safe, and developer-friendly systems language and extendr as the user-friendly bridge to R, analogous to Rcpp for C++. The speakers detail extendr's architecture, its growing adoption on CRAN, and showcase Rust's key advantages, including its helpful compiler, ease of refactoring, and simple path to writing powerful parallel code.

The session then transitions to practical applications. Wuji provides a detailed demonstration of building a bioinformatics package from scratch to calculate splicing indices, highlighting how extendr simplifies the management of complex dependencies (like hts-lib) and enhances performance. The event concludes with an extensive Q&A, tackling crucial topics such as the process for submitting Rust-based packages to CRAN and Bioconductor, the challenges of code review for a new language, and the technical requirements for build system toolchains.

Integrating Rust and R for Bioinformatics: A Bioconductor Developer Forum

• 0:00:01 Introduction: The Bioconductor Developer Forum kicks off, setting the agenda to discuss the implementation of Rust in R packages for bioinformatics, with a focus on the extendr framework and the Bioconductor ecosystem.
• 0:04:48 What is extendr?: The extendr project is introduced as a framework for extending R with the Rust programming language, serving as a modern, safe, and performant alternative to using C++ with Rcpp.
• 0:05:35 Why Use Rust?: Rust is presented as a low-level language praised for its exceptional performance, memory safety (enforced by its "borrow checker"), and a highly developer-friendly compiler that provides clear, actionable error messages, making it easier to learn and debug.
• 0:11:21 Key Advantages of Rust in R:
  • Performance: Even simple, non-optimized Rust code is exceptionally fast.
  • Effortless Parallelization: A standard vectorized function can be converted into a multi-threaded, parallel one with just one or two additional lines of code using libraries like Rayon.
  • Broader Reach: Core algorithms written in Rust can be easily exposed to other scientific computing languages like Python (via PyO3) and Julia, maximizing their impact.
• 0:18:01 Concrete Example - mdl Package: A Rust-based re-implementation of R's model.matrix function is shown to be about 10 times faster. The key takeaway is that it's written in 100% "safe" Rust, which allows non-expert maintainers to confidently modify the code without introducing memory errors.
• 0:24:07 Releasing Packages to CRAN and Bioconductor: The speakers address the practicalities of publishing. The process for CRAN is now relatively painless, as extendr automatically "vendors" dependencies (bundling all required source code), ensuring packages are self-contained, reproducible, and compliant with CRAN policies.
• 0:42:15 Live Demo - Building a Single-Cell Analysis Tool: Bioinformatics researcher Wuji demonstrates building an R package with a Rust backend to calculate Percent Spliced-In (PSI) values. Highlights include:
  • Simple Dependency Management: Adding a complex library like hts-lib (for reading genomic data files) requires only a single line in a configuration file, avoiding the complicated Makevars setup required with C++.
  • Interactive Prototyping: Rust functions can be written, compiled, and tested directly and interactively from an R session, speeding up development.
  • Long-Term Maintainability: Rust's strict type system helps document the code and makes it easier for the original author and others to understand and modify it in the future.
• 1:01:06 Community Building and Code Review: The forum discusses the logistics of incorporating Rust into the Bioconductor ecosystem, acknowledging the need for reviewers familiar with the language. To facilitate this, a dedicated #rust channel is created on the Bioconductor Zulip chat for ongoing discussion.