๐งฌ DoHealth
Understand a cancer. See it in 3D. Help fight it.
When someone you love is diagnosed with cancer, you feel helpless. DoHealth turns that helplessness into understanding โ and a way to help.
๐ Live demo ยท ๐ค Pitch slides ยท ๐ Built for the DigitalOcean Gradient AI Hackathon โ AI for Social Good
๐ The problem
Every day, families receive a diagnosis packed with terrifying words โ "KRAS G12C," "EGFR L858R" โ and no way to make sense of them. The tools that explain cancer at the molecular level (PyMOL, RCSB, AlphaFold) are built for experts who already know the protein, the mutation, and the drug.
Everyone else โ the parent, the partner, the friend โ is locked out of understanding the disease attacking someone they love. And they desperately want to help.
๐ก The solution
DoHealth makes molecular oncology understandable and participatory for anyone. Ask in plain English, get three things:
๐ง 1. Understand
GLM-5.2 explains the cancer, the mutation, and how the treatment works โ in warm, jargon-free language a scared family member can actually follow.
๐ฌ 2. Explore
The real protein target and the drug that binds it, in interactive 3D with the mutation highlighted โ pulled live from RCSB PDB, AlphaFold, and PubChem.
๐ค 3. Contribute
Fold a novel protein on a real GPU with ESMFold. From Folding@home to the Foldit game, ordinary people have advanced cancer research for years โ now anyone can take part.
โก Powered by DigitalOcean Gradient
| ๐ค Gradient serverless inference (GLM-5.2) | Turns plain language into the exact structural biology that drives the 3D scene |
| ๐ App Platform | Hosts the full-stack app with auto-deploy |
| ๐๏ธ GPU Droplets (RTX 6000 Ada, 48GB) | Live ESMFold protein folding |
๐ Data sources
All free, public, and open โ no API keys except Gradient:
| Source | What we pull |
|---|---|
| ๐งฌ RCSB PDB | Experimental protein 3D structures (by PDB ID, e.g. 6OIM) |
| ๐ฎ AlphaFold DB | Predicted structures (by UniProt ID) โ fallback when no experimental structure exists |
| ๐ PubChem | Small-molecule drug 3D structures (by name) |
๐๏ธ ESMFold (facebook/esmfold_v1) |
Folds novel sequences into new structures, on our GPU |
| โญ Curated hero set | KRAS ยท EGFR ยท BRAF ยท BCR-ABL reference cases |
โค๏ธ Honest by design
In-silico predictions are hypotheses for scientists to validate โ not treatments. DoHealth accelerates understanding and participation; it never diagnoses or prescribes. Handling a vulnerable audience responsibly isn't a limitation โ it's the point.
๐งฌ DoHealth โ because understanding is the first way to fight back.
Log in or sign up for Devpost to join the conversation.