AlphaFold3 Format¶
Expected input format for tsp-maker parse af3.
Directory Structure¶
af3_outputs/
├── P12345/ ← Folder name becomes protein ID
│ ├── P12345_summary_confidences.json
│ ├── P12345_confidences.json
│ ├── seed-1_sample-0/
│ │ ├── model.cif
│ │ ├── summary_confidences.json
│ │ └── confidences.json
│ ├── seed-1_sample-1/
│ │ └── ...
│ └── ...
├── Q67890/ ← Folder name becomes protein ID
│ └── ...
└── ...
Folder Names = Protein IDs
The folder name (e.g., P12345) becomes the protein ID in your dataset. Ensure folders are named with the identifiers you want. See Protein ID Rules.
Required Files¶
Per-Sample Files¶
Each seed-{X}_sample-{Y}/ directory must contain:
| File | Description |
|---|---|
model.cif |
Structure in mmCIF format |
summary_confidences.json |
Summary metrics |
confidences.json |
Full confidence data |
summary_confidences.json¶
{
"ranking_score": 0.85,
"ptm": 0.82,
"iptm": 0.78,
"fraction_disordered": 0.05,
"has_clash": false
}
confidences.json¶
Ranking¶
Samples are ranked by ranking_score from summary_confidences.json:
- All
seed-*_sample-*directories are scanned - Sorted by
ranking_score(descending) - Top N samples extracted
Extracted Metrics¶
| Metric | Source |
|---|---|
plddt_mean |
Mean of atom_plddts |
plddt_min |
Minimum |
plddt_median |
Median |
ptm |
summary_confidences |
iptm |
summary_confidences (multimers) |
ranking_score |
summary_confidences |
fraction_disordered |
summary_confidences |
has_clash |
summary_confidences |
pae_mean/min/max |
From PAE matrix |
Multimer Detection¶
Multimer predictions are detected by the presence of iptm in summary confidences.
Additional multimer metrics:
- chain_ptm
- chain_iptm
- chain_pair_iptm
Example Command¶
Output Naming¶
Files are named with _AF3_ suffix:
P12345_AF3_1.cifP12345_AF3.jsonP12345_AF3_1.npy
AlphaFold Server Output¶
The parser handles outputs from both: - Local AlphaFold3 installation - AlphaFold Server downloads
Both produce the same directory structure.