An apology first

This post should have been written a year ago. The two workshops it describes took place in August and September 2025, and a report in Japanese appeared in the newsletter of the Japanese Society for Bioinformatics (JSBi) in January 2026. The report to the wider, English-speaking Bioconductor community is only arriving now, in November 2026.

I am sorry for the delay. I owe this apology to the participants, helpers and co-hosts who made the workshops happen, and to everyone in the Bioconductor community who has been curious about how teaching Bioconductor in a local language went in Japan. Part of the reason for the delay is explained later in this post, because it is tied to the future of our translation infrastructure.

The first Bioconductor Carpentries workshops taught in Japanese

Bioconductor has been a member organisation of The Carpentries since August 2022 (announcement). The Bioconductor Training Committee maintains a set of Carpentries-style lesson modules, ranging from an introduction to R and Bioconductor to an advanced course on bulk RNA-seq analysis. How these lessons are developed and taught is described in Drnevich et al. (2025), PLOS Computational Biology.

Bringing this training to communities around the world in their own languages is supported by the Chan Zuckerberg Initiative (CZI) Essential Open Source Software for Science (EOSS) grant Delivering High Quality Bioconductor Training for a Worldwide Community.

As part of that effort, the Japanese Bioconductor community organised two two-day workshops in 2025, co-hosted with the Japanese Society for Bioinformatics (JSBi). To our knowledge these were the first Bioconductor workshops ever taught in Japanese. Both were hybrid events, with an in-person classroom and a Zoom connection for remote participants. I (Kozo Nishida) was the instructor for both. I serve on the Bioconductor Community Advisory Board and in the working group that translates Bioconductor teaching materials, and these workshops were the first time our Japanese translations were used in front of a class.

Where the materials live

Everything we used is in the bioconductor-translations GitHub organisation, and anyone is welcome to use it for their own teaching:

• Japanese lesson translations: bioc-intro-ja, bioc-project-ja and bioc-rnaseq-ja.
• Workshop websites and slides: 2025-08-05-osaka, 2025-08-05-osaka-slides, 2025-09-11-kyoto and 2025-09-11-kyoto-slides.
• The organisation also hosts the French translation bioc-intro-fr, which was used in the Bioconductor course in Benin.

Why this post is late, and what I have learned about our translation setup

Here is the honest part. One reason this post took so long is that I have not been able to maintain bioconductor.crowdin.com and the bioconductor-translations repositories the way they deserve, and I felt guilty writing a cheerful report about workshops while the infrastructure behind them was falling behind.

Some background. I was the one who proposed that the Bioconductor community use Crowdin to manage translations. The model was The Turing Way, which translates its handbook on Crowdin with great success. Crowdin’s support for open-source projects has been generous, and I remain very grateful for it.

However, the two projects differ in an important way. The Turing Way’s source documents are plain Markdown. The Carpentries lessons are written in R Markdown, with executable code chunks and lesson-specific fenced blocks for callouts, challenges and solutions. To feed them into Crowdin we had to convert R Markdown to Markdown into separate translation repositories, sync the translated strings back, and reassemble buildable lessons from the result. The Crowdin and GitHub integration that this requires is complex and fragile, and it is not easy to manage.

At the same time, the world has changed since I made that proposal. AI translation quality has improved substantially, and with Agent Skills an AI agent can now produce a first translation according to project rules: a shared glossary, consistent terminology, and clear instructions about what to leave untouched, such as code, package names and function names. The resulting translation can then be reviewed by community members with appropriate language and domain expertise. Under these conditions, managing translations directly in GitHub, with translated lessons living alongside their sources and reviewed through pull requests, is more efficient than keeping a Crowdin and GitHub integration alive.

So, in order to make Bioconductor teaching materials available in more languages, I no longer think we should keep using bioconductor.crowdin.com, and I intend to retire it. This is not a complaint about Crowdin. It is a recognition that the tool was the right one for a different kind of source document, and that a simpler workflow now serves our lessons better.

What comes next

I am setting up new, GitHub-only translation repositories together with an Agent Skill for rule-following AI translation, its documentation, and the outreach needed to bring more languages and more contributors on board. This work is supported by the Bioconductor Developer Champions Program.

Please give me a little more time before the formal announcement. When it is ready, I will share it here on the blog.

Get involved

• Teach with the materials. The Japanese lessons, the workshop websites and the slides linked above are openly licensed and free for anyone to use.
• If you are interested in translating Bioconductor teaching materials into your own language, watch the bioconductor-translations organisation for the upcoming announcement.
• Join the Bioconductor community through the support site and the community chat.

Acknowledgements

Thank you to Naohisa Goto and Atsushi Fukushima for helping at and hosting the workshops, to JSBi for co-hosting them, and to every participant. This work was supported by the Chan Zuckerberg Initiative EOSS program and by the Bioconductor Developer Champions Program.