lucidrains/chroma-pytorch — explained in plain English
Analysis updated 2026-07-20 · repo last pushed 2022-12-27
Design a protein that binds to a specific target like a virus spike protein.
Generate novel protein-based drug candidates digitally before lab testing.
Prototype custom proteins for new vaccines or materials research.
| lucidrains/chroma-pytorch | affaan-m/jarvis | cvlab-kaist/geometric-action-model | |
|---|---|---|---|
| Stars | 159 | 158 | 158 |
| Language | Python | Python | Python |
| Last pushed | 2022-12-27 | — | 2026-07-17 |
| Maintenance | Dormant | — | Active |
| Setup difficulty | hard | hard | hard |
| Complexity | 5/5 | 5/5 | 5/5 |
| Audience | researcher | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires PyTorch with GPU support and familiarity with protein structure data formats.
Chroma-pytorch is an open-source attempt to recreate a powerful AI system called Chroma, originally developed by the biotech company Generate Biomedicines. Chroma's purpose is to design entirely new proteins from scratch. Proteins are the molecular machines that drive biology, and being able to design them on demand has massive implications for creating new medicines, vaccines, and materials. Instead of waiting to discover useful proteins in nature, this software aims to generate them digitally. The system works using a type of artificial intelligence called a "generative model," which is similar to the technology behind AI image generators. It learns the complex 3D rules of how proteins fold and function by studying vast amounts of biological data. To create a new protein, it starts with random molecular noise and gradually refines it step-by-step into a stable, realistic protein structure. The project uses a combination of neural network techniques to understand the intricate spatial relationships between thousands of atoms. Researchers, biologists, and machine learning engineers would use this to design custom proteins for specific jobs. For example, as noted in the project's documentation, you could theoretically design a protein that binds directly to the coronavirus spike protein to neutralize it, or create entirely new protein-based drugs. It gives scientists a way to prototype biological solutions digitally before ever stepping into a physical lab. The project is an early-stage, work-in-progress effort to make advanced protein design accessible to the broader research community. The creator notes that this approach to protein design is an emerging field, with competing academic labs exploring similar methods. It is built in Pytorch, a popular programming tool for AI, and the developer plans to integrate additional scientific language models as the project evolves. The effort is tied to OpenBioML, a broader initiative to keep cutting-edge biological AI freely available to the public.
An open-source AI project that designs entirely new proteins from scratch using generative models, recreating a system originally built by Generate Biomedicines for digital protein design.
Mainly Python. The stack also includes Python, PyTorch.
Dormant — no commits in 2+ years (last push 2022-12-27).
Open-source project tied to the OpenBioML initiative, aimed at keeping cutting-edge biological AI freely available to the public.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
double-check against the repo, no cap.