Spaces:

---

pimenova

/

ProteinTTT

like 7

Running on Zero

Dear Hugging Face Team,
I am a researcher working in the groups of Josef Sivic at CTU Prague. Our recent work, "One Protein Is All You Need" has been accepted at ICLR 2026. In this paper, we introduced ProteinTTT, a method that enables self-supervised customization of protein language models to a single target protein on the fly.

These promising results open up the opportunity to apply our model to many other case studies with real impact. However, it may be difficult for our target users, researchers with a biological background, to set up the model for making predictions on their data. Therefore, we have implemented this Hugging Face Space, which provides a very convenient interface for non-ML experts. We have received positive feedback from the bio community, and currently, several independent labs are interested in using this Space.

We have a very positive track record with Hugging Face Spaces. Our team has developed several tools that are gaining significant traction in the community. DreamMS that has become a daily tool for many researchers. PPIformer, which focuses on protein-protein interaction modeling.

The fundamental limitation we face is that it is technically impossible to run our inference pipeline on a CPU as the backbone model ESMFold is optimized for NVIDIA kernels. A ZeroGPU grant would allow us to overcome this hardware barrier and provide the biological community with the real-time, high-accuracy structural predictions necessary for true scientific impact.

Thank you for considering our application!

Best regards,
Olga