Protein Structure · biohub
ESMFold2-Experimental-Fast
A fast, experimental protein structure prediction model generating all-atom coordinates from amino acid sequences.
What ESMFold2-Experimental-Fast is
ESMFold2-Experimental-Fast is a protein structure prediction model developed by biohub. It represents a speed-optimised variant within the experimental ESMFold2 branch, pairing accelerated inference with exploratory training modifications.
The model produces all-atom three-dimensional macromolecular coordinates in PDB format from input amino acid sequences. Alongside coordinate outputs, it reports confidence metrics including residue count, mean predicted local distance difference test (pLDDT) on a 0–100 scale, and predicted template modelling (pTM) scores.
This variant is released primarily to facilitate the reproducibility of findings detailed in biohub's corresponding research publication. For standard research applications, the developers recommend using the main ESMFold2 model.
How it works
ESMFold2 predicts high-resolution, all-atom three-dimensional protein structures. The architecture supports predictions conditioned either on a single sequence or on multiple sequence alignments (MSAs) to improve modelling accuracy on challenging targets.
The model is distributed in the safetensors format compatible with the transformers library and is executed via EvolutionaryScale's esm software package. Running inference requires 16 GB of GPU memory.
Key facts
- Task
- Protein structure prediction
- Input modality
- Protein sequence
- Output formats
- PDB coordinates, residue count, mean pLDDT, pTM
- GPU memory requirement
- 16 GB
What you can use it for
- Rapid all-atom structure generation
- Predicting three-dimensional all-atom coordinates in PDB format from an input sequence using accelerated inference.
- Fold confidence estimation
- Extracting global confidence scores, specifically residue count, mean pLDDT, and pTM, to evaluate model certainty.
- Study reproducibility
- Replicating structural benchmark figures and experimental evaluations published in the accompanying bioRxiv paper.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
ESMFold2-Experimental-Fast
Strengths and limits
Good at
- Combines experimental ESMFold2 research updates with accelerated inference speeds.
- Generates complete all-atom atomic coordinates in standard PDB format.
- Supplies both mean pLDDT and pTM metrics to quantify prediction confidence.
Watch out for
- Intended primarily for reproducibility of published research rather than routine research usage, for which the standard ESMFold2 model is advised.
- Requires a GPU with at least 16 GB of memory to run inference.
Inputs and outputs
You provide
Protein sequence
One-letter amino-acid codes.
You get back
All-atom coordinates in PDB format, viewable and downloadable.
Residue count, mean pLDDT (0–100) and pTM.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use ESMFold2-Experimental-Fast commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
How do I run ESMFold2-Experimental-Fast without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on CPU only, typically in 12s, and a run costs about 28 credits. Visitors without an account get 1,000 credits a month.
What inputs does ESMFold2-Experimental-Fast accept?
The model accepts an amino acid sequence provided as plain text.
What outputs are returned after running the model?
It outputs all-atom atomic coordinates in PDB format alongside a metrics table reporting the total residue count, mean pLDDT (scaled 0 to 100), and the pTM score.
Should this model be chosen over standard ESMFold2?
No; the developers advise using standard ESMFold2 for research investigations, noting that the experimental models exist to support replication of published findings.
Can ESMFold2 utilize multiple sequence alignments?
Yes, ESMFold2 supports structure prediction conditioned either on a single sequence or on multiple sequence alignments to assist with difficult prediction targets.
How to cite it
Written from ESMFold2-Experimental-Fast's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.