Dataset for: Predicting Experimental Success in De Novo Binder Design: A Meta-Analysis of 3,766 Experimentally Characterised Binders

Benchmarking dataset for the publication: Predicting Experimental Success in De Novo Binder Design: A Meta-Analysis of 3,766 Experimentally Characterised Binders
Inlcludes the following data: 

• final_data.csv: Dataset with all collected binder features with the following relevant columns
  • binder_id: name of binder
  • target_id: name of target
  • binder: binary experimental binding info
  • source: source publication of the binder
  • All other columns are the features described in the publication
• input_pdbs.tar.zst: Input .pdb files used for analysis where file names correspond to binder_id in the final_data.csv; sourced from public available de novo binder design campains
• AF2_initial_guess_outputs.tar.zstd: Output .pdb files generated with the initial guess AF2 implementation (https://github.com/nrbennet/dl_binder_design)
• AF3_outputs.tar.zstd: Compressed output from AF3 including structure (.cif) and confidence files (.json) with one subfolder per binder_id
• Boltz1_outputs.tar.zstd: Compressed output from Boltz-1 including structure (.cif), summary confidence (.json) and pae/plddt (.npz) files with one subfolder per binder_id
• ColabFold_outputs.tar.zstd: Compressed output from ColabFold including structure (.pdb) and confidedence files (.json)