orphan-disease-opportunity-mcp
# `orphan-disease-opportunity-mcp`
[](./LICENSE)
[](https://creativecommons.org/licenses/by/4.0/)
[-3776AB.svg)](./pyproject.toml)
[](https://rare-disease-atlas.github.io)
**Open Venture & Translational Opportunity Atlas — Model Context Protocol (MCP) Server**
**Flagship Deep-Dive #1:** `JAK2 V617F` Myelofibrosis (`jak2-myelofibrosis`) • **Web Dossier:** <https://rare-disease-atlas.github.io>
`mcp-name: io.github.rare-disease-atlas/orphan-disease-opportunity-mcp`
An open, **zero-third-party-dependency** Python MCP server and OpenAPI 3.1 toolkit designed for technical founders (**Computational / AI-Bio**, **Public Health Systems**, **Venture / Novel Financing**) and autonomous AI research agents evaluating high-tractability orphan disease startups, drug-repurposing wedges, non-dilutive SBIR/venture-philanthropy grants, and ethical clinical validation pathways.
---
## Instant 1-Command Installation (Zero PyPI Wait Required)
Run directly from GitHub via [`uvx`](https://docs.astral.sh/uv/) with **zero external runtime dependencies** (pure Python >= 3.10 standard library):
```bash
uvx --from git+https://github.com/rare-disease-atlas/orphan-disease-opportunity-mcp orphan-disease-opportunity-mcp
```
*(Once published to PyPI, `uvx orphan-disease-opportunity-mcp` also runs identically.)*
### Add to Claude Code (1 Command)
```bash
claude mcp add orphan-disease-opportunities -- uvx --from git+https://github.com/rare-disease-atlas/orphan-disease-opportunity-mcp orphan-disease-opportunity-mcp
```
### Add to Claude Desktop / Cursor / Windsurf / Cline / Antigravity
Add to `claude_desktop_config.json`, `.cursor/mcp.json`, or your MCP client settings:
```json
{
"mcpServers": {
"orphan-disease-opportunities": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/rare-disease-atlas/orphan-disease-opportunity-mcp",
"orphan-disease-opportunity-mcp"
]
}
}
}
```
---
## Why This Exists
Over 7,000 rare diseases carry **FDA Orphan Drug Act** structural advantages (7-year US market exclusivity, 100% waiver of the **$4,682,003 FY2026 PDUFA filing fee**, 25% qualified clinical trial tax credit, and N=120–350 pivotal trials), yet technical builders outside traditional wet-lab pharma rarely see where **single-mutation structural tractability** intersects **urgent standard-of-care failure**.
**Flagship Case Study #1 (`jak2-myelofibrosis`)** models this end-to-end:
- **Causal Precision:** A single somatic mutation (`JAK2 V617F` in the **JH2 pseudokinase** autoinhibitory domain) drives ~50–60% of Primary Myelofibrosis (`ORPHA:824`, ~15,000–25,000 US patients).
- **First-Gen SoC Failure Cliff:** All 4 FDA-approved JAK inhibitors (*ruxolitinib, fedratinib, pacritinib, momelotinib*) are Type I ATP-competitive binders of the conserved **JH1 catalytic domain**, inhibiting wild-type JAK2 alongside V617F—causing dose-limiting anemia/thrombocytopenia, ~50% discontinuation within 3 years, minimal clonal eradication, and **11–23 month median survival post-ruxolitinib failure**.
- **Zero-Wet-Lab & Low-Capital Wedges:** Structure-guided JH2 V617F vs WT (`PDB: 4FVR/6E2Q/8BAK/8EX2` vs `4FVP`) allosteric/PROTAC design + outsourced CRO screening; AI marrow fibrosis digital pathology (`MF-0..3` composite endpoints solving the Phase 3 SVR35/TSS50 crisis); single-cell megakaryocyte-stroma drug repurposing (`GSE144568`, `GSE156644`); no-washout RWE clinical trial matching; allo-HSCT timing CDS; and evergreen venture-philanthropy SPVs.
---
## Exposed MCP Capabilities
### Tools (`tools/list` & `tools/call`)
| Tool | Parameters | Capability |
| :--- | :--- | :--- |
| **`match_founder_opportunities`** | `background` (`computational-ai-bio`, `public-health`, `venture-novel-financing`, `all`), `capital_tier` (`bootstrapped-grant`, `seed`, `series-a-platform`, `all`), `wet_lab_required` (`true`/`false`), `disease_id` | Filters startup & translational blueprints with unit economics, regulatory pathways, open datasets, and 90-day milestones. Syncs live from `https://rare-disease-atlas.github.io/opportunities.json` with automatic fallback to bundled `opportunities.json`. |
| **`get_disease_case_study`** | `disease_id` (default `"jak2-myelofibrosis"`) | Returns causal JH2 vs JH1 biology, Type I JAKi failure modes, FDA Orphan Drug economics, $1.7B–$2.9B M&A comps, and open `PDB`/`GEO`/`ChEMBL`/`ClinicalTrials.gov` accessions. |
| **`find_non_dilutive_grants`** | `track` (`all`, `sbir-sttr`, `foundation`, `venture-philanthropy`, `regulatory-orphan`, `federal-translational`) | Maps NCI/NHLBI SBIR, MPN Research Foundation (`MPN Challenge`), Blood Cancer United (`TAP`), Harrington Discovery Institute, FDA OOPD, and ARPA-H awards. |
| **`get_validation_communities`** | `disease_id` (default `"jak2-myelofibrosis"`) | Maps institutional registries (`MPN PROGRESSion Registry NCT07362225`, `HealthTree`, `Blood Cancer United CTSC`, `MPN Voice`) paired with strict **non-extractive ethical patient engagement rules**. |
### Resource (`resources/read`)
- **`briefing://rare-disease-playbook/jak2-myelofibrosis`** — Complete Markdown venture & translational playbook.
### Prompt (`prompts/get`)
- **`evaluate-rare-disease-venture`** — Multi-step 90-day killer-experiment, non-dilutive grant-stacking, and ethical validation workflow (`founder_background`, `capital_budget`, `wet_lab_access`).
---
## Alternative Deployment Modes
### Docker (Hardened Non-Root Container)
```bash
docker build -t orphan-disease-opportunity-mcp .
# Stdio MCP transport
docker run -i --rm orphan-disease-opportunity-mcp
# Remote Streamable HTTP MCP server (exposes POST /mcp, GET /.well-known/mcp.json, GET /health)
docker run --rm -p 8080:8080 orphan-disease-opportunity-mcp --transport http --host 0.0.0.0 --port 8080
```
### Local Source Execution
```bash
git clone https://github.com/rare-disease-atlas/orphan-disease-opportunity-mcp.git
cd orphan-disease-opportunity-mcp
python3 -m orphan_disease_mcp
```
### Custom GPT Actions & OpenAPI 3.1 Agents
Import [`openapi.yaml`](./openapi.yaml) directly into **OpenAI Custom GPT Actions**, LangChain `OpenAPIToolkit`, or any OpenAPI 3.1 client to query `https://rare-disease-atlas.github.io/opportunities.json` over HTTPS with zero backend infrastructure.
---
## Environment Variables (Optional)
| Variable | Default | Description |
| :--- | :--- | :--- |
| `ATLAS_DATA_URL` | `https://rare-disease-atlas.github.io/opportunities.json` | Override URL for live JSON dataset sync. |
| `ATLAS_OFFLINE` | unset (`0`) | Set to `1` (`true`) to force deterministic offline execution from bundled `opportunities.json`. |
---
## Verification & Quality Assurance
Run the full zero-dependency end-to-end test suite (covers `stdio` JSON-RPC 2.0 lifecycle, all 4 tools, live HTTP mock sync + offline fallback, Streamable HTTP transport, and `server.json` / `glama.json` / `smithery.yaml` / `openapi.yaml` schema checks):
```bash
python3 test_mcp.py
```
## License
- **MCP Server Code:** [MIT License](./LICENSE) (`Copyright (c) 2026 Rare Disease Venture Atlas`)
- **Opportunity & Case-Study Dataset (`opportunities.json`):** [Creative Commons Attribution 4.0 International (`CC-BY-4.0`)](https://creativecommons.org/licenses/by/4.0/)
TDQS
Scored across 5 tools
Each tool targets a clearly distinct step in orphan-disease opportunity discovery: matching founders, retrieving disease case studies, finding grants, identifying validation communities, and submitting feedback. There is no meaningful overlap between the tools, and the descriptions make their boundaries explicit.
All tool names use consistent snake_case with a clear verb_object pattern such as match_founder_opportunities, get_disease_case_study, and find_non_dilutive_grants. The naming is predictable and readable throughout.
Five tools is well-scoped for this specialized workflow and each tool maps to a distinct need in the founder opportunity discovery process. None of the tools feels redundant or extraneous.
The surface covers the core lifecycle: opportunity matching, disease diligence, funding search, community validation, and feedback. Minor gaps remain, such as no explicit broad disease/opportunity search or deeper founder-blueprint retrieval separate from the matching tool, but agents can work around these through the existing tools.