DISCO-design

DISCO-design / DISCO

Public

Code for the DISCO model: General Multimodal Protein Design Enables DNA-Encoding of Chemistry

disco-design.github.io diffusion-models enzyme-design protein-design
47
4
100% credibility
Found Apr 09, 2026 at 47 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

DISCO generates protein sequences and 3D structures that bind to specified small molecules, DNA, or RNA using diffusion-based AI.

How It Works

1
🔬 Discover DISCO

You hear about DISCO, a tool that designs custom proteins to bind any molecule like drugs or DNA.

2
📥 Get it ready

Download and set up DISCO on your computer with a simple install command.

3
📝 Describe your target

Create a short list in a text file telling DISCO what molecule or DNA you want proteins for, like a chemical formula.

4
🚀 Generate designs

Run one command and watch DISCO create dozens of protein ideas that fit perfectly around your target.

5
📁 Review results

Open the folder to see protein blueprints as 3D models and matching sequences ready to build.

✅ Ready for lab

Pick your favorite design and bring it to life in experiments or simulations.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 47 to 47 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is DISCO?

DISCO is a Python-based generative model for joint protein sequence and 3D structure design, conditioned on ligands via SMILES or files, DNA/RNA sequences, or nothing for unconditional generation. It outputs foldable PDBs and FASTAs from simple JSON inputs—no templates or inverse folding needed. Run via CLI like `python runner/inference.py experiment=designable input_json_path=your_config.json seeds=[0,1,2]`, with presets for fast prototyping or max quality.

Why is it gaining traction?

Hits SOTA co-designability on 178/179 ligands in the Studio-179 benchmark, plus DNA/RNA binders and de novo enzymes like heme carbene transferases. Effort levels (fast/max) trade speed for quality, resumable jobs skip duplicates, and Hugging Face hosts models/data. Stands out from pipelines by letting sequence objectives guide structures directly.

Who should use this?

Protein engineers designing ligand binders or nucleic acid interfaces without scaffolds. Synthetic biologists prototyping evolvable enzymes from reactive intermediates. Computational chemists validating hits via in silico screening before lab work.

Verdict

Solid starter for code github ai in protein design—strong README with examples, uv install, and arXiv backing—but 47 stars and 1.0% credibility signal early maturity; test on your targets before production. Worth forking the code github repo if bio-AI fits your stack.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.