Frequently in scientific research, we must write code in order to perform our desired analysis. Whether we are writing a short script for our own manuscript, or making an app we intend to share with other researchers, incorporating standard programming practices will enhance the readability and reproducibility of our work. Below, we provide an overview of what coding languages are commonly used at the Hutch, different resources for running code, and recommendations for developing research software.
One of the most commonly asked questions when researchers realize they need to use some form of code in their work is, “Which language should I learn?” This simple question rarely has a simple answer, and thus we aim to provide both resources for training opportunities as well as basic documentation about what other people commonly use. Your results may vary and we highly recommend you get started by taking some basic introductory courses on common languages either in person or on the web in order for you to better evaluate your next steps.
There are several questions you can ask that help you figure out what programming language to learn first:
- What are other people in your immediate work group using? If you’ll need to use or adapt code from other people, it makes sense to start with the tool they’re using.
- What tools are most commonly used in your field? You may find assistance answering this question by reviewing recently published literature, or through talking with colleagues (on Twitter, at conferences, etc).
- What are your career goals? Are the tools you’re planning to learn going to be useful only for your current position, or will they be meaningful additions to your resume?
- What do you personally prefer? Some people simply prefer working in one language over another, because the syntax (rules about the structure of the language) makes more sense to them.
For a clear side-by-side comparison of R and Python, this article provides some general considerations from a data science perspective.
It’s important to be deliberate in your choice of first language to learn, because it can be a significant time investment, but it’s not committing you to only using that language in the future. Just like learning human languages, learning one language generally makes it easier to learn another language later.
Getting Started with Common Languages
If you are new to coding, or are interested in learning about coding languages for research software, the Fred Hutch offers training resources for many of the common languages and tools listed below. For more information on training opportunities, cooperative learning groups, and how to get started, please see the reference page.
The remainder of the Software Development content will presume a basic skill level in at least one common language, as well as a basic understanding of GitHub for version control.
R is a programming language and also a software development environment. It is widely used among statisticians and has strong capabilities for statistical modeling and data analysis. While R’s core functions are fairly small, there is a robust community of user-contributed R ‘packages’ (eg. see “Bioconductor” on the target reference page). You can download R for your computer, install it from Center IT’s Self Service (on Macs or on PCs), or run R on SciComp’s computing clusters using environment modules.
RStudio is a graphical frontend to R that also improves upon a basic R installation, providing syntax-highlighting and code-completion, static or dynamic reports (via RMarkdown documents), and easing the creation of R packages, among other functionalities. It is considered an IDE which functions much like a wrapper around R itself, to create a graphic user interface, and easy access to various tools and functions that enhance the user’s experience of using R.
Python is another language used extensively within the bioinformatic community. A very high-level comparison of Python to other commonly used languages is that it’s generally on the easier side for being able to learn and understand, but it doesn’t give you as much detailed control over the details of computation compared to C. In other words, it’s easier to use, but not quite as performant. One of the nice things about Python in recent years has been that there is a large community of software developers contributing highly efficient code as installable modules, which makes the entire codebase more valuable for your average user.
Unix is the foundation for both Linux and macOS, and is the operating system that is most commonly used for developing and executing bioinformatic software tools. In order to navigate a Unix-based operating system and execute commands, it is extremely useful to use the command line interface, which is generally referred to as BASH (Bourne-Again SHell).
While version control software has evolved over time, a new evolution that is happening more and more is the need for a wider group of researchers to actively use version control resources to manage their code and documentation of processes that are ongoing in their research. From the perspective of reproducibility, shareability and interoperability, the need for a sharing platform that integrates version control and collaboration is becoming more and more a critical part of a researcher’s toolkit. Thus, regardless of the degree to which code plays a direct role in a research project, more and more often at least a cursory understanding of what GitHub is and how it can be utilized in scientific research is important.
You have the option to run code on a local (desktop) computer, or on a shared computing cluster. This section helps you figure out what will work best for your task and help you make sense of the Large Scale Computing section if you do need more computing resources than your local computer.
Other Languages and Tools for Data Analysis
As we said, while there are some commonly used languages, there really are a huge variety of software people use for research. This section includes some introductions to other types of software for which we do not specifically have an article addressing them. Resources external to this Wiki or specific Resource Library entries are linked to where applicable. Happy (coding) Trails!!!!
Proprietary (Licensed) Statistics Software
We encourage the use of free, open-source software at the Hutch, especially for academic research use. For projects and researchers who prefer not to use open-source software, Fred Hutch supports proprietary software through shared licenses for the following statistical analysis and visualization software: SAS, SPSS, Stata, and Matlab users. Please see the Resource Library for information on accessing Matlab on Fred Hutch compute resources. For more information on obtaining access to licensed software, please contact Center IT.
There is an active Tableau data visualization community at Fred Hutch. For more information on help sessions, meetups, and Q&A, please see the #tableau-user-comm channel on The Coop Communities Slack.
Workflow managers are software that facilitate the setup, execution, and monitoring of scientific analyses involving multiple steps. These are especially important for complex bioinformatic analyses that require various software dependencies and resource requirements, and can make your analysis both easier to troubleshoot and more reproducible. Workflow managers common for scientific processes include: Snakemake, Cromwell, Toil, CWL, and Nextflow. For more information on running Nextflow, please see this demo.
Nextflow and Cromwell are currently under development at the Fred Hutch for wider, supported use by researchers here. Nextflow is currently available and there is a growing community group working collaboratively you can join on their Slack channel, and the same is available for Cromwell including a Slack channel.
The Julia language aims to combine the accessible syntax of R or Python with the speed of C/C++ programs. While not currently as functional as R or Python for bioinformatic tasks, there is a growing collection of resources for Julia for bioinformatics.
Updated: October 6, 2023Edit this Page via GitHub Comment by Filing an Issue Have Questions? Ask them here.