Skip to content

Bioconductor-to-Galaxy auto-integration

The load-bearing GBCC2025 storyline — the joint Galaxy-Bioconductor event's central rationale. A pair of oral talks plus a workshop plus a CoFest deliverable establish the pipeline for wrapping Bioconductor R packages as Galaxy tools, with GSVA as the proof-of-concept and an AI-driven automation pipeline as the production target.

  • Cross-platform integration: a pair of GBCC2025 talks + a workshop + a CoFest deliverable that together establish the pipeline for wrapping Bioconductor R packages as Galaxy tools
  • Lead community: Cleveland Clinic / Galaxy Project (Daniel Blankenberg group) + Bioconductor maintainers
  • CoFest deliverable: GSVA Galaxy tool wrapper (draft PR) — GSVA on Bioconductor (v2.6.1, maintainer Robert Castelo)
  • Status at GBCC2025: integration / infrastructure (multi-talk + workshop + CoFest)

Sessions at GBCC2025

Talk 1 — Day 2 Oral Session 1 (#3)

  • Speakers: Jayadev Joshi, Fabio Cumbo, Daniel Blankenberg (Cleveland Clinic / Galaxy)
  • Title: "Automated integration of R/Bioconductor tools into Galaxy"
  • Theme: the production pipeline — taking arbitrary Bioconductor packages and emitting Galaxy tool XML + Conda environment + test data automatically

Talk 2 — Day 2 Oral Session 2A (#8)

  • Speakers: Fabio Cumbo, Jayadev Joshi, Daniel Blankenberg
  • Title: "Towards the automatic integration of tools in Galaxy: the anvi'o suite and a new AI-driven automation pipeline"
  • Theme: AI-driven extension of the same pipeline — using LLMs to infer parameter schemas, generate test cases, and reduce the manual effort of Galaxy tool wrapping. anvi'o is the case study; Bioc is the broader target

Workshop — Day 3

  • Presenter: Daniel Blankenberg (Cleveland Clinic / Galaxy Project)
  • Title: "Streamlining Bioinformatics: Adding Bioconductor Tools in Galaxy"
  • Format: hands-on; participants wrap a Bioconductor package as a Galaxy tool live

CoFest deliverable — Day 5-6

  • Target: GSVA (Bioc) → Galaxy tool wrapper, draft PR
  • Team: Fabio Cumbo, Jayadev Joshi, Bryan Raubenolt, Daniel Blankenberg + on-site Sean Doyle, Polina Yagudayeva, Charlotte Soneson, Marcel Ramos Pérez; remote Robert Castelo (GSVA maintainer)
  • Why GSVA: widely-used pathway-scoring method, stable API, Castelo is engaged as remote maintainer — low-risk proof-of-concept for the auto-integration pipeline

What it does (and why it matters)

Bioconductor's value is its 2,000+ curated R packages; Galaxy's value is reproducible web-accessible workflow execution for users who don't write code. The historical friction is that wrapping a Bioc package as a Galaxy tool is hand-built, version-fragile, and labor-bottlenecked on a small number of maintainers. The GBCC2025 sequence — two talks scoping the automation pipeline, a workshop teaching the manual baseline, a CoFest sprint producing a real wrapper (GSVA) — is the joint community's coordinated push to industrialize the bridge: an AI-assisted generator that takes a Bioc package's DESCRIPTION + man pages + vignette and emits a working Galaxy tool with minimal human review.

How it works

Core idea. Auto-translate the metadata that already exists in an R/Bioconductor package — DESCRIPTION, roxygen-documented function signatures, man pages, vignette code — into the four artefacts a Galaxy tool needs: a tool XML (parameters + outputs + help), a Conda <requirements> block pinning R + Bioc + the package, a wrapper R script that maps Galaxy CLI args back to the function call, and a test definition with reference inputs.

Inputs / outputs. Input: a target R/Bioconductor package (URL or name) and the chosen entry-point function(s). Output: a PR-ready Galaxy tool wrapper — XML + R driver + Conda env + tests — that can be merged into galaxyproject/tools-iuc and from there into IWC workflows.

Key innovation. LLM-assisted wrapper drafting (Cumbo / Joshi / Blankenberg). The initial GSVA wrapper was drafted by Gemini from the package's R-side documentation, then refined by humans — proof that LLM agents can produce the tedious schema-translation step. A parallel Python-side pipeline (the R2G2 framework and the anvi'o argparse-driven approach) shows the same principle ports to Python tools.

Parameters / API surface worth knowing. - Source signal — R DESCRIPTION (package + version + deps), roxygen @param tags (parameter types + help), vignette code (canonical usage examples for test fixtures). - Output artefacts — Galaxy tool XML (<inputs>, <outputs>, <command>, <requirements>, <tests>); R wrapper script; Conda env file; reference test data. - LLM step — used to map roxygen @param descriptions to Galaxy XML parameter types (boolean / integer / float / select / data) and to draft help text. - Manual review gate — humans verify the generated XML, parameter coverage, and test outputs before merging the PR.

Canonical example. GSVA (the CoFest deliverable). The pipeline ingested the GSVA Bioconductor package, produced a draft Galaxy tool wrapper exposing the method, kernel, and min.sz / max.sz parameters, ran the wrapper against a reference expression matrix + gene-set collection, and confirmed the output gene-set-by-sample score matrix matched a direct R call — making GSVA the demonstration that the auto-wrap pipeline can produce a real, mergeable Galaxy tool from a real Bioc package.

Where it fits in the corpus

  • AACR 2026: axis = bioinfo-tools (Galaxy-wrapped Bioc tools are exactly what enables wet-lab cancer biologists to run reproducible analyses without an R bench scientist)
  • mia (entry) — Tuomas Borman is core to both the microbiome Bioc ecosystem AND its Galaxy bridge; the GBCC mia framing emphasized Galaxy / cross-platform export
  • GAIA (entry) and GalaxyMCP (entry) — sister AI-/automation-flavored Galaxy talks; together with this entry they trace the AI-augmented future of the Galaxy platform
  • atena (entry) — Castelo-maintained; a natural next package for the same auto-integration pipeline once GSVA is shaken out
  • EuroBioC 2025: not presented (this is the GBCC-specific story — the joint event's whole rationale)
  • Nextflow Summit 2026: parallel question in a different ecosystem (Nextflow vs. Galaxy as workflow substrates); not directly cross-referenced but worth tracking

Notes

This is the load-bearing GBCC story — the "joint Galaxy-Bioconductor conference" exists as a venue precisely to coordinate this integration work. The CoFest GSVA wrapper is the proof-of-concept that the automation pipeline can produce a real, mergeable Galaxy tool from a real Bioc package; the auto-integration pipeline (talks 1 + 2) is the production target that scales the proof-of-concept to the long tail of Bioc. If the pipeline ships, the bridge between Bioconductor and Galaxy stops being a per-package craft project.