RFantibody
Design new antibody and nanobody CDRs from an antigen structure and an antibody framework, proceeding through backbone design, sequence design, and structure filtering.
API/CLI tool ID: rfantibody
Understanding inputs and outputs
The antigen is the target recognized by the antibody; its structure goes in targetPdb. frameworkPdb contains the antibody framework that supports the regions around the CDRs. The backbone is the protein’s common structural scaffold, while the sequence is the order of its amino acids. numDesigns sets the number of backbone candidates, and seqsPerStruct sets sequences per backbone; they count different things.
temperature controls diversity during sequence sampling; it is not an experimental temperature. pred_lddt is structural confidence on a 0–1 scale. interaction_pae is the predicted error in the relative placement of antibody and antigen, while framework_aligned_cdr_rmsd measures CDR displacement after aligning the frameworks. Small errors or displacements do not establish actual binding or antibody function.
Input settings
| Setting | Format | Required | Default | Description and constraints |
|---|---|---|---|---|
targetPdb | pdb | Required | — | Experimental antigen (target) structure — Extensions pdb · cif |
frameworkPdb | pdb | Required | — | Antibody framework structure (the official framework from the repository's design campaign is provided) — Extensions pdb · cif |
numDesigns | number | Optional | 10 | Number of backbone designs to generate (rfdiffusion -n, 1–100). For more than 100, contact us at /contact?topic=pricing with the number you need and your input size. — Minimum 1, Maximum 100 |
temperature | number | Optional | 0.1 | ProteinMPNN sampling temperature (documented default: 0.1) — Minimum 0, Maximum 1 |
seqsPerStruct | number | Optional | 4 | Sequences to design per structure (proteinmpnn -n) — Minimum 1, Maximum 64 |
numRecycles | number | Optional | 10 | RF2 recycling steps (documented default: 10) — Minimum 1, Maximum 30 |
deterministic | boolean | Optional | false | Reproducible run (--deterministic) |
seed | number | Optional | — | Random seed for reproducibility (rf2 -s) |
Output metrics
| Column | Display name | Range | Unit | Description |
|---|---|---|---|---|
pred_lddt | RF2 pLDDT (0–1) | 0~1 | — | Structure confidence from the RF2 filter. Not affinity or the probability of experimental success. |
interaction_pae | Interface PAE | — | Å | Predicted aligned error at the antibody–antigen interface |
pae | PAE | — | Å | Predicted aligned error of the whole complex |
framework_aligned_cdr_rmsd | CDR RMSD | — | Å | CDR displacement after aligning on the framework |
See Reading results for metric definitions and interpretation. A range shows possible values; whether higher or lower is better depends on the metric.
Execution availability
Open an execution plan in your project to see available tool versions and costs. Review and approve the inputs, settings, and terms before running.
License
Code, model weights, and accompanying data have their own terms. The table summarizes terms from official sources; check your execution plan for the configuration you will use. License texts are available in Third-party notices.
| Component | License | Commercial use | Source | Notes |
|---|---|---|---|---|
| Code | MIT | Allowed, subject to notice requirements | License text | — |
| Model weights | MIT | Allowed, subject to notice requirements | Official terms and distribution information | The RFdiffusion_Ab, ProteinMPNN_v48_noise_0.2, and RF2_ab weights may be used under the MIT license, as confirmed in the official repository. Copyright and permission notices must be retained. |
| Model weights | CC-BY-4.0 | Conditional; see notes | Weight distribution metadata | ADAPT immune-receptor design parameters (RFab_noframework-nosidechains-5-10-23_trainingparamsadded.pt) use CC BY 4.0 and require attribution. |
The pipeline includes RFdiffusion, ProteinMPNN, and an antibody-tuned RoseTTAFold2 filter; PyRosetta is not required. RF2 code is bundled in the repository. The NVIDIA base container permits hosted services, but its distribution conditions also apply to internet services. Do not expose NVIDIA APIs directly; provide substantial additional functionality and comply with protective terms, notices, and separate component licenses.
Checked on 2026-11-09.