Magellan

Inspiration

Magellan is inspired by a blunt reality: rare disease patients are stuck waiting because building new drugs from scratch is too slow and too expensive for tiny populations, while pharma is sitting on a graveyard of shelved, post Phase I or Phase II assets and a huge universe of generics and patent expiring drugs that could be repurposed if anyone could systematically identify the best matches.

What it does

Magellan is a drug to disease and disease to drug matching platform for rare diseases. You input a drug or a rare disease, Magellan ingests the relevant clinical, mechanistic, and safety data, normalizes everything into a shared mechanism space, then produces a ranked shortlist with an explanation for each match, including the mechanistic nodes involved, evidence links, safety flags, and an explicit uncertainty score.

How we built it

Magellan is a Python 3.11, FastAPI, Celery, SQLAlchemy backend running in Docker, using PostgreSQL and Redis, and integrating with public biomedical APIs like PubChem, ChEMBL, ClinicalTrials.gov, PubMed/NCBI, ClinVar, openFDA, and Orphanet to build explainable mechanism‑level drug–disease scores and evidence.

Challenges we ran into

Our biggest challenge was normalization and defensible scoring with incomplete data. We cannot rank drug to disease matches unless both entities are canonicalized, mapped to consistent pathway nodes, and backed by an evidence ledger that can survive scrutiny. The second hard part was forcing uncertainty to be first class.

What we learned

Rare disease repurposing is hindered by a lack of systematic prioritization, missing data structure, and the absence of decision support that ties biology, evidence, and regulatory feasibility into a single workflow. We also learned that explainability is not a nice-to-have. It is the product. If a foundation, KOL, or BD team cannot see exactly why a drug ranked highly, they will not trust it, and will not get access to assets or data.

What's next for Magellan

Ultimately, the long-term ambition is to transform how society thinks about abandoned or generic drugs. Instead of being sunk cost artifacts, they become continuously mined therapeutic options for small populations that traditional pharma economics ignores. If successful, Magellan would help close the gap between the thousands of known rare diseases and the small fraction that currently have approved therapies by making repurposing systematic, defensible, and financially rational.

Built With

• alembic
• and-integrating-with-public-biomedical-apis-like-pubchem
• celery
• chembl
• clinicaltrials.gov
• clinvar
• compose
• fastapi
• ncbi
• openfda
• postgresql
• psycopg2
• pubchem
• pubmed
• pubmed/ncbi
• python
• pyyaml
• redis
• requests
• sqlalchemy
• sqlalchemy-backend-running-in-docker
• using-postgresql-and-redis
• uvicorn