Instructions to use biohub/ESMFold2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use biohub/ESMFold2 with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("biohub/ESMFold2", dtype="auto") - Notebooks
- Google Colab
- Kaggle
Binder design with ESMFold2 - Live webinar with the Biohub Research team
Binder design with ESMFold2
Join us for a live webinar with the Biohub Research team
Tuesday, June 30, 2026, 1 PM PT / 4 PM ET.
Register Now: https://biohub.zoom.us/webinar/register/6317817963570/WN_wNh50zuzR76u6nhuR_-WMg
ESMFold2 turns ESMC's sequence representations into molecular models of biomolecular complexes. We used it to design minibinders and scFvs against five cancer and immunology targets in days.
Lab validation confirmed hits across all targets with nanomolar affinity, epitope specificity, selectivity over close homologs, and functional activity in a cell-based immune checkpoint assay β and one of our predictions has been validated with a cryo-EM structure.
Since launch we have kept going. The open source protocol is now faster, runs on smaller GPUs, and is easy to run on our notebook.
Sign up now for the webinar to:
- Hear directly from the researchers who led this work
- Get a detailed walk-through of the protocol and latest improvements
- Come away with the ability to design and run it on your own targets
Our research team will also be available at the end for a Q&A, so bring your questions.