{"id":51197088,"url":"https://github.com/bioMate-AI/biomate-bioconductor-kb","last_synced_at":"2026-07-16T06:00:47.487Z","repository":{"id":363211852,"uuid":"1250860168","full_name":"bioMate-AI/biomate-bioconductor-kb","owner":"bioMate-AI","description":"BioMate-KB Bioconductor Skills — 200 packages (top 100 by downloads + 100 rising stars) as vignette-grounded Claude/agent skills, with per-package workflow recipes","archived":false,"fork":false,"pushed_at":"2026-06-15T23:21:08.000Z","size":1210,"stargazers_count":352,"open_issues_count":0,"forks_count":40,"subscribers_count":22,"default_branch":"main","last_synced_at":"2026-06-16T01:11:46.896Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/bioMate-AI.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":"CITATION.cff","codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-27T03:10:25.000Z","updated_at":"2026-06-16T00:09:43.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/bioMate-AI/biomate-bioconductor-kb","commit_stats":null,"previous_names":["biomate-ai/biomate-bioconductor-kb"],"tags_count":1,"template":false,"template_full_name":null,"purl":"pkg:github/bioMate-AI/biomate-bioconductor-kb","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bioMate-AI%2Fbiomate-bioconductor-kb","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bioMate-AI%2Fbiomate-bioconductor-kb/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bioMate-AI%2Fbiomate-bioconductor-kb/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bioMate-AI%2Fbiomate-bioconductor-kb/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/bioMate-AI","download_url":"https://codeload.github.com/bioMate-AI/biomate-bioconductor-kb/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bioMate-AI%2Fbiomate-bioconductor-kb/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35532646,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-16T02:00:06.687Z","response_time":83,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-06-27T21:31:06.299Z","updated_at":"2026-07-16T06:00:47.474Z","avatar_url":"https://github.com/bioMate-AI.png","language":"Python","funding_links":[],"categories":["Industry Skills"],"sub_categories":[],"readme":"# BioMate-KB — Bioconductor Skills\n\n**200 Bioconductor packages — the top 100 most-downloaded plus the top 100 rising stars — formatted as Claude Code Skills.**\n\nA skill bundle in [Claude Code Skills](https://docs.anthropic.com/en/docs/claude-code/skills-overview) format. It covers the 100 most-used Bioconductor packages **and** the top 100 rising-star packages (newly released since ~2021 and fast-growing) from the BioMate-KB knowledge base. Each skill teaches Claude when to choose a package, **the workflows it supports** (each analysis as a recipe), what parameters to set, how to interpret results, and what pitfalls to avoid.\n\n## What's here\n\n```\nskills/  (200 packages · 12 domains)                  ⭐ = rising star\n├── transcriptomics/  (97 — DESeq2, edgeR, limma  ·  ⭐ standR, Voyager, sechm, crisprScore)\n├── genomics/         (33 — GenomicRanges, Biostrings  ·  ⭐ rBLAST, syntenet, ggmanh)\n├── general/          (19 — Biobase, DOSE  ·  ⭐ immunotation, faers, mosbi)\n├── proteomics/       (16 — MSnbase, mixOmics, mzR  ·  ⭐ MatrixQCvis, MsDataHub, TargetDecoy)\n├── epigenomics/      (10 — ChIPseeker, minfi  ·  ⭐ HiCExperiment, HiContacts, epigraHMM)\n├── single-cell/      (8  — SingleCellExperiment  ·  ⭐ demuxmix, hoodscanR, MuData)\n├── variant-calling/  (4  — VariantAnnotation, snpStats, vsn)\n├── metagenomics/     (4  — phyloseq, microbiome, DirichletMultinomial)\n├── imaging/          (4  — flowCore, EBImage  ·  ⭐ lisaClust, cytoviewer)\n├── annotation/       (2  — biomaRt, KEGGgraph)\n├── enrichment/       (2  — enrichplot, ReactomePA)\n└── metabolomics/     (1  — ⭐ rgoslin)\n```\n\nA package that supports multiple analyses lists each as a `### ` recipe under a **`## Workflows`** section (e.g. DESeq2 → standard / multi-factor / LRT; crisprscore → on-target / off-target / indel scoring). The most-used 100 cover ~56% of Bioconductor analysis-package download volume; the rising-star 100 surface newly-important methods before they reach the top by volume. Beyond this 200-package sample, the full BioMate KB has runnable workflows for **1,818** Bioconductor packages — **~87% of analysis-package download volume** — available via [BioMate Cloud](https://biomate.ai).\n\n**Bioconductor version:** these skills are grounded against **Bioconductor 3.21** (pinned explicitly — `release` is a moving pointer that drops packages as it advances, e.g. several rising stars dropped out of 3.23). The pinned version is recorded in `MANIFEST.json` (`bioconductor_version`); re-fetch a different snapshot with `BIOC_VERSION=… python3 extraction/fetch_authoritative_sources.py`.\n\n## Full package list\n\nAll **200 packages** — with category, description, and the workflows each one supports — are in\n**[PACKAGES.md](PACKAGES.md)** (one page, four columns; ⭐ marks rising stars).\n\n### Packages per domain\n\n![Packages per domain](assets/packages_per_domain.png)\n\n### Workflows per domain\n\nA package with multiple analyses contributes one `### ` recipe subsection per workflow — **390 workflows\nacross the 200 packages; 89 are multi-workflow** (e.g. DESeq2 → DE / multi-factor / QC-transform /\nLRT; crisprScore → 6 scoring recipes).\n\n![Workflows per domain](assets/workflows_per_domain.png)\n\n## How these 200 were selected\n\nTwo ranked sets of 100:\n\n- **Top 100 by downloads** — ranked purely by the official Bioconductor download score\n  (`bioc_pkg_scores.tab`). These cover **~56% of Bioconductor analysis-package download volume**. Because the rank\n  is by raw volume, this set includes both **analysis tools** (DESeq2, edgeR, limma, fgsea, …) and\n  the **foundational data-structure / I/O / annotation** packages nearly every analysis imports\n  (GenomicRanges, Biostrings, SingleCellExperiment, AnnotationHub, …).\n- **Top 100 rising stars** — restricted to **analysis** packages (infrastructure, data-container,\n  and GUI packages excluded), recently released (first release ≥ 2021) and fast-growing in 2025\n  (year-over-year download growth + ≥ 3,000 distinct download IPs), ranked by 2025 downloads. They\n  surface newly-important methods *before* they reach the top by raw volume.\n\nTogether these 200 packages cover **~57% of Bioconductor's analysis-package download volume**\n(foundational infrastructure and data-container packages — which alone are ~40% of raw\ndownloads — are not counted as analysis volume). The full **BioMate-KB** goes much further:\nrunnable workflows for **1,818** analysis packages — **~87% of analysis-package download\nvolume**.\n\n\u003e **Note on domains.** Domain labels come from BioMate's catalog and are intentionally coarse —\n\u003e *transcriptomics* is a broad catch-all that absorbs most single-cell, spatial, and gene-set tools\n\u003e (scater / scran / monocle / SingleR are single-cell; fgsea / GSVA are enrichment) — which is why\n\u003e it dominates the charts, and why the small *single-cell* (8) and *annotation* (2) folders are\n\u003e remnants of the same imperfect classifier rather than clean boundaries. We keep BioMate's labels\n\u003e for traceability; treat the domain folders as a rough guide, not a strict ontology.\n\n**Our contribution** — unique among public skill libraries:\n\n1. **The only per-R/Bioconductor-package library** — **200** packages here, **1,818** in the full KB.\n   Every other public set is task-oriented or Python-centric; none provide per-package R/Bioconductor\n   knowledge at this depth (the closest R one, wolf5996, has 13 skills and covers R *packaging*, not\n   Bioconductor analysis).\n2. **Vignette-grounded \u0026 fact-verified** — every R function named in a skill is verified to appear in\n   that package's own Bioconductor vignette (mean verify **0.91**), not free-form LLM prose.\n3. **Per-package workflow recipes** — a package's distinct analyses are explicit `### ` recipes\n   (**390** across the 200) — *how to run it*, not just *what it is*.\n4. **Executable backing** — the same knowledge runs as managed, validated workflows on\n   [BioMate Cloud](https://www.biomate.ai) for **1,818** packages.\n\n\n## Tutorial: Bulk RNA-seq → DESeq2 → STRINGdb Gene Interaction Network\n\nThis walkthrough shows how BioMate routes a plain-English bulk RNA-seq request through DESeq2, GO enrichment, and an **interactive STRINGdb protein-protein interaction panel** — without writing R code.\n\n**What the demo covers:**\n- Natural-language workflow routing to DESeq2 + clusterProfiler\n- Differential expression on 120 samples (human GRCh38, control vs treated)\n- GO enrichment (enrichGO) identifying apoptotic regulation as the top pathway\n- **Interactive gene interaction panel** — STRINGdb network visualization of the top 2,841 DEGs, with nodes coloured by log₂FC and edges weighted by STRING confidence score\n- AI findings summary + citable methods section + citation export\n\n[![BioMate AI | Bulk RNA-seq: DESeq2 + GO enrichment, no R code needed](https://img.youtube.com/vi/WL8Jk7M8n1g/maxresdefault.jpg)](https://youtu.be/WL8Jk7M8n1g)\n\n*▶ Click the image above to watch the tutorial on YouTube.*\n\nFor all tutorials (single-cell Seurat, DNA methylation, bispecific antibody design, CAR-T, base editing, GLP-1 modality selection, BCMA myeloma triage) see the **[BioMate Tutorial Page →](https://biomate.ai/tutorials.html)**\n\n---\n\n## Want the full collection?\n\nThis bundle covers 200 Bioconductor packages (top 100 by downloads + 100 rising stars). **[BioMate AI](https://www.biomate.ai)** gives you:\n\n- **Broad Bioconductor coverage** — runnable workflows for **1,818** packages (**~87%** of analysis-package download volume), plus nf-core and drug-discovery pipelines across genomics, proteomics, and more\n- **Efficient parallel computing** — workflows run in the cloud with automatic scaling, no cluster setup required\n- **Interactive visualization \u0026 analysis** — inspect, filter, and re-run results through linked charts and per-step QC dashboards, with AI-assisted interpretation that links every claim back to the underlying data\n- **Reproducible reporting** — methods and results documents generated with complete parameter and software-version provenance, formatted for publication and audit\n\n**Free for academic and non-profit researchers** — [register at www.biomate.ai](https://www.biomate.ai) (no credit card required).\nCommercial plans available. Questions or collaboration inquiries: [contact@biomate.ai](mailto:contact@biomate.ai)\n\n## How to use\n\n```bash\n# Clone\ngit clone https://github.com/bioMate-AI/biomate-bioconductor-kb.git\ncd biomate-bioconductor-kb\n\n# Install all skills into Claude Code (global)\nfind skills -name \"SKILL.md\" | while read f; do\n  pkg=$(dirname \"$f\" | xargs basename)\n  cp \"$f\" ~/.claude/skills/bioconductor-${pkg}.md\ndone\n```\n\nOr copy a single domain:\n\n```bash\n# Only RNA-seq DE skills\nfind skills/transcriptomics -name \"SKILL.md\" | while read f; do\n  pkg=$(dirname \"$f\" | xargs basename)\n  cp \"$f\" ~/.claude/skills/bioconductor-${pkg}.md\ndone\n```\n\nEach `SKILL.md` is a self-contained Claude Code skill file — Claude discovers it automatically once it's in `~/.claude/skills/` (global) or `.claude/skills/` (project-level).\n\n## Ranking source\n\nPackages are ordered by Bioconductor's official monthly download score:\n- Source: \u003chttps://bioconductor.org/packages/stats/bioc/bioc_pkg_scores.tab\u003e\n- Snapshot taken: 2026-05-21\n- Top 100 by download score + 100 analysis rising stars = 200 packages, together **~57% of Bioconductor analysis-package download volume**\n\nBecause the ranking is by download volume, the bundle includes both **analysis tools** (DESeq2, edgeR, limma, fgsea, …) and the **core data-structure, I/O, and annotation packages** that nearly every analysis depends on (GenomicRanges, Biostrings, SingleCellExperiment, AnnotationHub, …) — the latter rank highly precisely because they are imported everywhere. Both are useful to an agent: the analysis packages teach *how to analyze*, the foundational ones *how to represent and load* the data.\n\n## Knowledge layer, not pipelines\n\nThese skills are the **knowledge layer** — when and why to use each package, with parameters, assumptions, pitfalls, and alternatives. They are not runnable pipelines and carry no infrastructure details.\n\n**BioMate hosts and executes these workflows for you** — managed compute, automated QC, and reproducible outputs — powered by this same Bioconductor know-how. For end-to-end cloud execution, see **[BioMate](https://www.biomate.ai)**.\n\n## License\n\n- **Skill content** (`skills/**/*.md`, `MANIFEST.json`): **CC-BY-4.0** — share + adapt with attribution.\n- **Extraction scripts** (`extraction/*.py`): **Apache-2.0** — use, modify, distribute.\n- **Underlying Bioconductor packages** retain their own (mostly Artistic-2.0 / GPL) licenses.\n\n## Citation\n\nIf you use this skill bundle in research, please cite:\n\n\u003e Zhang, Y. (2026). *BioMate-KB: A Real-Execution-Validated Workflow Knowledge Base for Bioconductor* (3.0). Zenodo. https://doi.org/10.5281/zenodo.20616356\n\nAnd, for the execution-grounding methodology:\n\n\u003e Zhang, Y. (2026). *Structure Grounding Is Not Enough: Real Execution as the Ground Truth for LLM-Generated Bioinformatics Workflows* (Version v3). Zenodo. https://doi.org/10.5281/zenodo.20616544\n\nConcept DOIs (always resolve to the latest version): BioMate-KB — https://doi.org/10.5281/zenodo.20616355 · Structure Grounding — https://doi.org/10.5281/zenodo.20616543\n\n## Regenerating the bundle\n\n```bash\n# Re-fetch the latest Bioconductor download scores\ncurl -O https://bioconductor.org/packages/stats/bioc/bioc_pkg_scores.tab\n\n# Regenerate the download-ranked set (top 100 by default).\n# The 100 rising-star skills are a separately-curated set (analysis packages,\n# first release \u003e= 2021, ranked by 2025 download growth) generated per-package\n# via extract_skill.py, then vignette-grounded with enrich_v2_grounded.py.\npython3 extraction/generate_bundle.py --top 100\n\n# Single package\npython3 extraction/extract_skill.py \\\\\n    --db \u003cpath-to-biomate-knowledge-db\u003e \\\\\n    --pkg DESeq2 \\\\\n    --out my-deseq2-skill.md\n```\n\nThe extraction code (`extraction/extract_skill.py`) is intentionally minimal (~300 lines) and reads only from BioMate's public knowledge fields — `tool_knowledge.{use_cases, limitations, alternatives, recommended_parameters, primary_citation, benchmark_papers}` and `tools.scientific_context`. Internal and infrastructure fields are excluded.\n\n## Versioning\n\n**v2.0.0** (2026-06-15) — full history in **[CHANGELOG.md](CHANGELOG.md)**. Skills are pinned to\n**Bioconductor 3.21** (recorded in `MANIFEST.json`).\n\nSince v1.0.0: coverage doubled to **200 packages** (added the top 100 rising stars), per-package\n`## Workflows` recipes, and a full package table + coverage charts.\n\nPlanned: track new Bioconductor releases, expand coverage (top-1000 if community demand justifies),\nand refine `SKILL.md` sections (Q\u0026A, gotchas, more examples).\n\n## Contributing\n\nOpen an issue or PR for:\n- Errors in any SKILL.md\n- Suggestions for new sections to extract\n- Packages missing from the top-100 / rising-star sets that should be included\n\n## Acknowledgments\n\nBioconductor download statistics published by the Bioconductor Core Team. SKILL.md format from [Anthropic Claude Code](https://docs.anthropic.com/en/docs/claude-code/skills-overview).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FbioMate-AI%2Fbiomate-bioconductor-kb","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FbioMate-AI%2Fbiomate-bioconductor-kb","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FbioMate-AI%2Fbiomate-bioconductor-kb/lists"}