Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/reactome/reactome-curator-workflows


https://github.com/reactome/reactome-curator-workflows

Last synced: about 1 month ago
JSON representation

Awesome Lists containing this project

README 

          
# reactome-curator-workflows



Shared Claude Code workflow skills for Reactome curation and release operations.



## What This Is



This repository contains Claude Code skills — markdown-based instruction files that

Claude Code reads directly. Any curator with Claude Code installed can clone this repo

and immediately run any skill. No build process, no configuration beyond cloning.



Skills are added incrementally as we identify and vet repeatable workflows.

See `CLAUDE.md` for full project context.



---



## Documentation



Full setup and usage instructions are in:



**[Reactome_CuratorWorkflows_ClaudeCode_Guide_v1_3.docx](./Reactome_CuratorWorkflows_ClaudeCode_Guide_v1_3.docx)**



This covers one-time prerequisites, cloning the repository, running each skill, and

how to add new skills. Start here if you are setting up for the first time.



---



## Setup (Quick Reference)



### 1. Prerequisites (one-time)



Install Git: https://git-scm.com/downloads



Install Claude Code:



```bash

curl -fsSL https://claude.ai/install.sh | bash

```



Set your Anthropic API key (add to `~/.zshrc`):



```bash

export ANTHROPIC_API_KEY=your-key-here

```



Install Python dependencies (use `pip3` on macOS):



```bash

pip3 install pandas openpyxl

```



### 2. Clone the repository



```bash

git clone https://github.com/reactome/reactome-curator-workflows.git ~/Developer/reactome-curator-workflows

cd ~/Developer/reactome-curator-workflows

```



### 3. Run Claude Code



```bash

cd ~/Developer/reactome-curator-workflows

claude

```



Claude Code automatically reads `CLAUDE.md` and loads all skills at session start.

The `.claude/settings.json` file allowlists external hosts (e.g., NCBI E-utilities)

required by certain skills.



### 4. Keep up to date



```bash

cd ~/Developer/reactome-curator-workflows

git pull

```



---



## Available Skills



### `/internal-module-review`



Formal structured internal review of a Reactome pathway report against Curator Guide

V94 standards. Produces a prioritized review DOCX with seven sections. Upload the

pathway report DOCX and Curator Guide PDF to the conversation before invoking.

Optional modifiers: `disease`, `drug`, `large`.



**Sections:**



| Section | What it covers |

|---|---|

| 1 | Reaction connectivity, including Input→Output entity chain verification |

| 2 | GO Biological Process assignments |

| 3 | Literature references |

| 4 | Grammar, clarity, and summation quality |

| 5 | General curation quality |

| 6 | Prioritized issue summary (consolidated table, all sections) |

| 7 | Entity and event name conventions |



Section 7 checks every entity and event name against the Reactome controlled vocabulary

(Jupe et al. 2014) and associated naming rules. Five subsections cover EWAS/peptide names

(gene symbol, coordinates, PTM prefixes), reaction/event names (all 11 CV classes), small

molecule names, complex names (colon separator, square bracket hierarchy notation), and

set names (comma separator, candidate member notation). Naming convention reference files

are automatically loaded — no additional uploads required.



```

/internal-module-review "HHV8 Infection" R-HSA-9521541 "Lisa Matthews" 2026-04-15

```



---



### `/annotate-pathway-from-reviews-or-topic_name`



AI-assisted pre-curation workflow (v1.3, April 2026). Proposes a complete Reactome

pathway hierarchy — pathway names, subpathway names, and reaction names — and verifies

primary literature citations using a mandatory ten-step PMID verification protocol.



Requires claude.ai Projects (Pro/Team/Enterprise) or the Claude API. PubMed and PMC

MCP servers are recommended for live PMID verification and full-text fetch.



**Two entry modes:**



- **Mode A** — Curator supplies references (PMIDs, DOIs, or uploaded PDFs); assistant

  reads the papers and proceeds directly to Phase 1.

- **Mode B** — Curator supplies a biological topic; assistant performs live web searches

  for primary literature, presents a ranked candidate reference list, and waits for

  curator confirmation before proceeding.



**Session workflow:**



```

Session opening

    │

    ├── Mode A: Curator supplies references

    │       └── Assistant reads papers → proceeds to Phase 1

    │

    └── Mode B: Curator supplies a topic

            │   Assistant searches → presents ranked reference list

            └── STOP — curator confirms or modifies list

                    └── Curator replies "proceed"



Phase 1 — Hierarchy proposal

    Proposed pathway / subpathway / reaction names

    Evidence species flags (human / non-human / chimeric)

    Non-human and chimeric reactions paired with inferred human reactions

    Proceeds immediately to Phase 2 — no approval gate



Phase 2 — Literature verification

    Ten-step PMID verification protocol per reaction

    PMC full-text fetch and verbatim evidence sentence extraction

    Standard or inferred-human citation blocks

    Subpathway by subpathway — no stops between subpathways



STOP — full output presented for curator review

```



**Citation block formats:**



Standard (direct or indirect evidence):

```

Reaction name:    [name from hierarchy]

Primary PMID:     https://pubmed.ncbi.nlm.nih.gov/[PMID]/

Primary PMC URL:  https://pmc.ncbi.nlm.nih.gov/articles/PMCXXXXXXX/

Authors/Year:     [First author et al., Year, Journal]

Primary evidence: "[verbatim sentence including method, proteins, cell type, figure]"

Evidence type:    DIRECT / INDIRECT / OVEREXP ONLY

Species:          [human / murine / other]

Flags:            [mismatches, gaps, curator actions required]

```



Inferred human reaction:

```

Reaction name:    [inferred human reaction name]

Evidence type:    INFERRED FROM NON-HUMAN / INFERRED FROM CHIMERIC

Species:          Homo sapiens (inferred)

inferredFrom:     [paired non-human or chimeric reaction name]

Primary PMID:     https://pubmed.ncbi.nlm.nih.gov/[PMID]/

Authors/Year:     [same as paired evidence reaction]

Flags:            No independent human experimental evidence. Curator should

                  confirm human orthology is supported before submission.

```



**Species and chimeric reaction logic:**



| Experimental situation | Reaction type | isChimeric | Inferred human required |

|---|---|---|---|

| Human proteins/cells only | Human reaction | FALSE | No |

| Single non-human species only | Non-human reaction | FALSE | Yes |

| ≥2 species mixed in same assay | Chimeric reaction | TRUE | Yes |



Non-human proteins use initial capitalisation (Jak2); human proteins use ALL-CAPS (JAK2).



**Evidence classification:**



| Class | Reaction type | Examples |

|---|---|---|

| DIRECT | Reaction | co-IP, in vitro reconstitution, SPR/ITC, NMR, crystallography, direct enzymatic assay |

| INDIRECT | BlackBoxEvent (provisional) | KO, rescue, overexpression, inhibitor treatment, domain deletion |

| INSUFFICIENT | Do not annotate | Microarray alone, bulk proteomics alone, ChIP-seq alone, computational prediction alone |



**What is out of scope** — handled separately by the curator using database tooling:

Neo4j/gk_central queries, OLS ontology verification, UniProt REST API verification,

GtoPdb drug checks, full annotation table population, disease flag resolution, and

pathway diagram drawing. All such fields are marked **PENDING CURATOR VERIFICATION**.



**Key limitations:**



- **PMID fabrication risk.** The ten-step protocol catches fabricated PMIDs, but

  curators should independently confirm all PMIDs before database submission.

- **Full-text access.** PMC full text fetched for open-access papers only; paywalled

  papers fall back to web search and are flagged if unavailable.

- **Knowledge cutoff.** Mode B performs live searches, but curators should verify that

  the most recent relevant primary papers have been captured.

- **Ontology accessions.** GO, ChEBI, MONDO, and UniProt accessions are not resolved

  in this workflow — all marked PENDING CURATOR VERIFICATION.



For the full workflow documentation, citation block reference, species/chimeric rules,

and version history, see

`.claude/skills/annotate-pathway-from-reviews-or-topic_name/README.md`.



```

/annotate-pathway-from-reviews-or-topic_name

```



---



### `/generate-doi-batch`



Generates a CrossRef DOI batch XML file for a release using DOIs.xlsx. Requires

DOIs.xlsx locally (from Team Drive) and Python 3 + pandas.



```

/generate-doi-batch V94

```



---



### `/extract-reactions`



Extracts a reaction graph for a named pathway from one or more medical/biology review

PDFs. Writes two artefacts:



- `_reactions.csv` — columns: Title, Input, Output, Catalyst,

  Regulators, Reviews, Source1–Source5

- `_references.html` — every cited primary reference as a link



Reviews holds pipe-delimited supplied-PDF basenames. Each Source cell holds exactly

one URL chosen by a PubMed → PMC → DOI → publisher ladder; any 6th or further sources

are dropped. Pre-curation draft, not a curated entry.



```

/extract-reactions

```



---



### `/update-gdrive-readme`



Regenerates the Reactome Team Drive README as a richly formatted Google Doc. On every

run: clears the existing doc, walks the live Team Drive folder inventory via the Drive

API, and rewrites the doc with styled headings, linked folder tables, a contacts table,

and a live folder inventory tree.



Requires Python 3 and the Google API client libraries. Uses OAuth user credentials by

default — credentials are stored at `~/.config/reactome/` on each user's local machine

and are never committed to the repo. Supports `--dry-run` (preview, no write) and

`--depth N` (inventory depth, default 2).



```

/update-gdrive-readme

/update-gdrive-readme --dry-run

/update-gdrive-readme --depth 3

```



---



### `/reactome-qa-tracker`



Runs `compare_dirs.sh` against two QA output directories (old version vs. new

slice), filters out developer-only and backlog sections using a curator-vetted

skip list, presents a per-category intermediate review for approval, and produces

a polished multi-sheet `.xlsx` curator tracker with Status dropdowns

(`Not Done` / `Fixed` / `Skipped`), color-coded fill, and a Comments column on

every data row.



Can also start from an existing comparison file — supply a Google Sheets URL or

an uploaded `.xlsx` / `.csv` instead of running the script.



**Workflow:**



1. Provide `OLD_DIR` and `NEW_DIR` — version labels are taken from the directory

   basenames. Handles directory alias mapping (`command-line-runner` →

   `commandlinerunner`, `diagram-converter` → `diagram-qa`).

2. `compare_dirs.sh` (included in the skill directory) runs once per QA tool

   subdirectory; output is captured to `/tmp/_comparison.txt`.

3. **Intermediate review** — the skill pauses and presents a summary of included

   vs. skipped files per tool category and waits for curator confirmation before

   building the workbook.

4. Skip rules remove developer-only files and backlogs. Curator-actionable files

   retained by default: `Mandatory_Attributes`, `DT701-ReactionParticipantsMismatch`,

   `GT037-LiteratureReferenceRelationshipDuplication`,

   `GT090-CatalystActivityCompartmentDoesNotMatchReactionCompartment`,

   `Entities_With_CoV_Species_Without_Corresponding_Disease`,

   `ReactionlikeEvent_Regulations_Not_Reviewed`, `Review_Status`.

5. Output: `_QA_Curator_Tracker.xlsx` — one sheet per tool group

   plus an Overview with a status legend and sheet directory.



```

/reactome-qa-tracker

```



---



## Prerequisites by Skill



| Skill | Requirements |

|---|---|

| `/internal-module-review` | Internet access; pathway report DOCX and Curator Guide PDF uploaded to conversation |

| `/annotate-pathway-from-reviews-or-topic_name` | claude.ai Pro/Team/Enterprise (Projects) or Claude API; PubMed and PMC MCP servers recommended; internet access for Mode B |

| `/generate-doi-batch` | DOIs.xlsx from Team Drive; Python 3 with `pandas` and `openpyxl` |

| `/extract-reactions` | One or more review-article PDFs (absolute paths); internet access to `eutils.ncbi.nlm.nih.gov` |

| `/update-gdrive-readme` | Python 3; `pip install google-auth google-auth-oauthlib google-auth-httplib2 google-api-python-client`; OAuth credentials at `~/.config/reactome/credentials.json` |

| `/reactome-qa-tracker` | `compare_dirs.sh` (in skill directory); Python 3 with `openpyxl`; two QA output directories (or an existing comparison file) |



> **Note on `/extract-reactions` and NCBI access:** In Claude Code, `.claude/settings.json`

> allowlists `eutils.ncbi.nlm.nih.gov` automatically. In claude.ai (browser), add it manually

> via **Settings → Capabilities → Domain allowlist**. Without it, PMID resolution falls through

> to DOI URLs or blanks — PMIDs are never recovered from training data.



---



## Chrome Extensions



### `AICurator`



Install from https://chromewebstore.google.com/detail/aicurator/jkdcmghlpgfilhdlngmopljpocpbnhdc



Go to a fresh Google Sheet then open the extension:



1. **Main** — pick or create a project (a folder under

   `/aicurator/`) bound to a Google Sheet, configure the AI

   provider, manage settings.

2. **Extract** — feed review-article PDFs and a pathway name to an LLM,

   resolve PMIDs against PubMed, write a 12-column reaction table to the

   sheet.

3. **Summate** — for each row, send the cited PMID-prefixed PDFs and the

   row context to an LLM, draft a Reactome-style summation paragraph,

   write it to column B.

4. **Canonize** — replace protein/gene mentions in columns A–F with their

   canonical UniProt-confirmed (human-only, reviewed-first) gene symbols.



---



## Using Skills in the claude.ai Desktop App



1. Locate your skill folder: `.claude/skills//`

2. ZIP it: `zip -r skill-name.zip skill-name/`

3. Go to **Customize → Skills → Upload ZIP** in the Desktop app



---



## Contributing a Skill



1. Create a directory under `.claude/skills/` named for your skill

2. Add `SKILL.md` with YAML frontmatter (`name`, `description`) and instructions

3. Add any supporting files (templates, scripts, reference docs)

4. Test it locally with Claude Code

5. Open a PR with a brief description of what the skill does and when to use it



For detailed instructions on writing and committing a new skill, see the full guide

(`Reactome_CuratorWorkflows_ClaudeCode_Guide_v1_3.docx`).



Suggestions for improving `annotate-pathway-from-reviews-or-topic_name` — including

changes to the system prompt or the RLE annotation reference file — are welcome via

GitHub Issues. When proposing changes to the reference file, cite the specific section

of the Curator Guide or Data Model Glossary that supports the change, and note the

guide version.



---



## Repository Structure



```

reactome-curator-workflows/

├── CLAUDE.md                                              ← project context for Claude Code

├── README.md                                              ← this file

├── Reactome_CuratorWorkflows_ClaudeCode_Guide_v1_3.docx  ← full setup guide

├── .gitignore

└── .claude/

    ├── settings.json                                      ← Claude Code host allowlist

    └── skills/

        ├── internal-module-review/

        │   ├── SKILL.md

        │   ├── Reactome_InternalReview_PROMPT_v1_4.docx

        │   ├── Reactome_InternalReview_TEMPLATE.docx

        │   ├── Claude_ReactomeDataModelGlossary_V95_Final.docx.pdf

        │   ├── Curator_Guide_V94.pdf

        │   ├── EWAS_name_rules.docx                       ← EWAS naming rules (Section 7)

        │   ├── Rules_for_automatic_reaction_typing.docx   ← reaction/event naming rules (Section 7)

        │   ├── bau060.pdf                                 ← Jupe et al. 2014 CV paper (Section 7)

        │   ├── ptm_lookup.xlsx                            ← PSI-MOD → PTM prefix table (source)

        │   ├── ptm_prefixes.md                            ← PTM prefix table, plain text (auto-loaded)

        │   └── Small_molecule_renaming.xlsx               ← canonical small molecule names (Section 7)

        ├── annotate-pathway-from-reviews-or-topic_name/

        │   ├── SKILL.md

        │   ├── README.md

        │   └── Reactome_RLE_Annotation_Reference_V94.md

        ├── extract-reactions/

        │   └── SKILL.md

        ├── generate-doi-batch/

        │   ├── SKILL.md

        │   └── generate_crossref_xml.py

        ├── update-gdrive-readme/

        │   ├── SKILL.md

        │   └── update_drive_readme.py

        └── reactome-qa-tracker/

            ├── SKILL.md

            └── compare_dirs.sh

```



---



## Contact



Repo maintainer: Marc Gillespie (NYU / SJU) — gillespm@stjohns.edu

For curation standards questions: Reactome curator Slack channel

For repo/skill questions: open a GitHub issue