# Data definitions

- Target chain: A. Designed binder chain: B. Fragment numbering follows the supplied model.
- `id`: stable record identifier used in website links. `displayName`: public binder name.
- `targetSequence` / `binderSequence`: amino-acid sequences in the model.
- `iptm`: AF3 interface predicted TM score. `ipae`: minimum inter-chain PAE, Å.
- `irmsd`: interface RMSD between design and evaluation poses, Å.
- `monomerRmsd`: binder monomer RMSD from its AF2 comparison, Å.
- `rgRatio`: binder Cα radius of gyration divided by 2.2 × length^0.38.
- Structure catalogue criteria: iPTM ≥0.7, iPAE ≤1.5 Å, iRMSD ≤2 Å, monomer RMSD ≤2 Å and Rg ratio ≤1.25.
- Success rate: designs passing the first three criteria divided by generated designs; catalogue binder counts additionally use the monomer RMSD and Rg criteria.
- `delta`: complex minus sequence-matched apo target-interface pLDDT. Contacts use a 5 Å heavy-atom cutoff and equal-weight residue heavy-atom means. Null means no valid comparison is available.
- Catalogue `plddt`: per-residue target Cα pLDDT, 0–100.
- Confidence-file `plddt`: per-residue non-hydrogen-atom mean pLDDT, rounded to six decimal places. `plddt_ca`: original Cα values.
- `token_chain_ids`, `token_res_ids`, `residue_names`: chain, residue number and name for each confidence-array entry.
- `pae`: full, unaltered asymmetric PAE matrix in Å; row i is the alignment residue and column j the scored residue. PAE is not a distance or contact matrix. No apo PAE is available.
- AF3 generation timestamps are omitted from model metadata; coordinates, residue identifiers and original notices are unchanged. Confidence-file `model_sha256` and `source_structure_sha256` identify the original CIF before this metadata-only change; `SHA256SUMS` identifies the distributed files. Reduced confidence files omit unused atom arrays and contact probabilities.
- DisProt regions use model residue coordinates, with sequence-verified fragment or isoform mappings where needed. Disorder content is the fraction of model residues covered by the assigned disorder regions; unmapped residues are not classified as ordered.

All binding candidates and displayed complexes are computational predictions.

- Catalogue `targetPlddt` (Target pLDDT): equal-weight mean of residue Cα pLDDT over the sequence-matched apo target. Null means no matching apo model is available.