About AntiFold

AntiFold predicts sequences which fit into an input antibody variable domain structure. The tool outputs residue log-likelihoods in CSV format, and can sample sequences to a FASTA format directly. Sampled sequences show high structural agreement with experimental structures.

AntiFold is based on the ESM-IF1 model and is fine-tuned on solved and predicted antibody structures from SAbDab and OAS. For larger-scale use, the tool is freely available on GitHub.


Correspondence: opig[at]stats.ox.ac.uk

>   Run AntiFold

Please provide a valid antibody PDB file or SAbDab entry ID along with valid heavy and light chain labels for your job.

Residues in the antibody chains of your input PDB will be automatically renumbered using ANARCI with the IMGT numbering scheme. The IMGT-annotated PDB will be available for download as part of your job results.

The results from an example job with 6Y1L can be found here. You can also load an example job for testing by clicking on the right-hand-side button.

Input PDB file:
SAbDab PDB entry (Optional):
Heavy chain regions:
Light chain regions:
Heavy chain identifier:
Light chain identifier:
No. sequences to generate:
Temperature:

Magnus Haraldson Høie & Alissa Hummer et al. “AntiFold:
Improved antibody structure-based design using inverse folding” 2024. [ link]

Download BibTex Reference

Header image credit: David Goodsell