FoldCraft

Fold-conditioned de novo binder design

FoldCraft enables fold-conditioning of binder structure, enabling design of binders with diverse folds like TIM-barrels, solenoid folds or Ig-like domains. Using VHH conditioned framework FoldCraft can succesfully design single domain nanobody binders against diverse tergets. The FoldCraft pipeline is described in this preprint

You can easily run FoldCraft on Google Colab using this link. Or use this notebook, if you want to design VHH fold binders specifically

Paper: https://www.biorxiv.org/content/10.1101/2025.07.02.662497v1.abstract

Installation

First you need to install FoldCraft repository on your local machine:

git clone https://github.com/KhondamirRustamov/FoldCraft

Then run code below to download all requirements, ColabDesign and AlphaFold2 weights

bash install_foldcraft.sh --cuda '12.4' --pkg_manager 'conda'

NOTE: AlphaFold3, which has been used in the manuscript for VHH design benchmarking, should be installed separately as described in official repository

Running FoldCraft locally

To run FoldCraft locally you will need the following files:

--output_folder       ->      Folder to save the results
--binder_template     ->      Path to the binder template PDB file (required)
--target_template     ->      Path to the target template PDB file (required)
--target_hotspots     ->      Residue ranges for target hotspots, e.g., "14-30,80-81,90-102" (required)

Then you should activate you conda/mamba ebvionment with conda activate FoldCraft and run FoldCraft with following comand:

python FoldCraft.py \
      --output_folder design_1qys_pd_l1 \
      --binder_template 1qys.pdb \
      --target_template pd_l1.pdb \
      --target_hotspots '36-41,84-88,92-96' \
      --binder_hotspots '25-40,50-65' \
      --num_designs 40

Or run this code if you want to design nanobody fold-conditioned binders using VHH framework. In that case you don't need to specify the binder template pdb

python FoldCraft.py \
      --vhh \
      --output_folder design_vhh_pd_l1 \
      --target_template pd_l1.pdb \
      --target_hotspots '36-41,84-88,92-96' \
      --num_designs 40

You can also specify the ProteinMPNN optimization strategy, and ProteinMPNN weights to use, as well as other settings:

--vhh                 ->      Whether to use VHH framework to construct target cmap (all binder information would be ignored in that case)
--sample              ->      Whether to generate designs until the target number of successful designs is reached (default: False)
--target_success      ->      Target number of successful designs to generate - used only if --sample is enabled (default: 100)
--num_designs         ->      Number of design trajectories to generate - ignored if --sample is enabled (default: 1)
--binder_hotspots     ->      Residue ranges for binder, e.g. "14-30,80-81,90-102"
--binder_mask         ->      Residue ranges in the binder to mask (ignored during loss computation), e.g. "14-30"
--binder_chain        ->      Binder template chain (default: 'A')
--target_chain        ->      Target template chain (default: 'A')
--design_stages       ->      Number of each design stages in 3stage_design (default: 100,100,20)
--mpnn_weight         ->      PoteinMPNN weights to use: 'soluble', 'original' (default: 'soluble')
--redesign_method     ->      ProteinMPNN redesign strategy: 'full' or 'non-interface' (default: 'non-interface')
--mpnn_samples        ->      Number of sequences to sample with ProteinMPNN (default: 5)
--mpnn_backbone_noise ->      Backbone noise during sampling (default: 0.0)
--mpnn_sampling_temp  ->      Sampling temperature for amino acids 0.0-1.0 (default: 0.1)
--mpnn_save           ->      Whether to save MPNN sampled sequences

Credits

This repository uses code from:

• Sergey Ovchinnikov's ColabDesign (https://github.com/sokrypton/ColabDesign)

• Justas Dauparas's ProteinMPNN (https://github.com/dauparas/ProteinMPNN)

• Martin Pacesa's BindCraft (https://github.com/martinpacesa/BindCraft)