Protein Design · Meta FAIR
ESM-IF1
Generates protein sequences for specified backbone structures using ESM2 language model representations.
What ESM-IF1 is
ESM-IF1 is an inverse folding model developed by Meta FAIR for protein design tasks. Given an input protein backbone coordinate structure, it designs compatible amino acid sequences.
Unlike graph network architectures such as ProteinMPNN, ESM-IF1 is informed by ESM2 language model representations. Comparing its designs with ProteinMPNN outputs allows researchers to identify consensus sequences and improve confidence prior to experimental synthesis.
How it works
The model designs sequences for a target backbone coordinate structure using ESM2 representations rather than relying on a graph neural network. Generation is controlled by user-defined sampling temperatures and sequence counts, producing candidate sequences scored by perplexity and recovery.
Key facts
- Version
- 1.0
- Task
- Inverse folding
- Input modality
- Protein structure
- Hardware requirement
- 16 GB GPU memory
What you can use it for
- Protein backbone sequence design
- Generates viable amino acid sequences tailored to a specific structural chain provided in PDB format.
- Consensus sequence generation
- Runs alongside ProteinMPNN to find overlapping sequence predictions, boosting design confidence prior to synthesis.
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
ESM-IF1
Strengths and limits
Good at
- Offers a language-model-based formulation for inverse folding as an alternative to graph networks.
- Provides sequence recovery and model perplexity scores alongside generated candidates.
- Can be run alongside ProteinMPNN to derive consensus designs.
Watch out for
- Requires at least 16 GB of GPU memory for execution.
- Requires defined backbone structural coordinates in PDB format.
Inputs and outputs
You provide
Backbone (PDB)
Chainoptional
Which chain to design. Defaults to the first in the file.
Temperatureoptional
Low converges on the native sequence; high explores.
Sequencesoptional
You get back
Sequences with model perplexity and recovery.
Licence
Commercial use permitted · MIT
The MIT licence allows use in commercial research and products.
Questions
Can I use ESM-IF1 commercially?
Commercial use permitted. The MIT licence allows use in commercial research and products.
How do I run ESM-IF1 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 1× A100 16GB, typically in 9.0s, and a run costs about 56 credits. Visitors without an account get 1,000 credits a month.
What inputs does ESM-IF1 require?
The model takes a protein backbone file in PDB format, the specific chain identifier, a sampling temperature, and the number of sequences to generate.
What metrics does the model return with designed sequences?
ESM-IF1 returns generated sequences in a tabular format accompanied by model perplexity and sequence recovery metrics.
How does ESM-IF1 differ from ProteinMPNN?
ESM-IF1 relies on ESM2 language model representations rather than graph neural networks to design sequences for target backbones.
Written from ESM-IF1's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.