Antibody design

What is antibody design?

Antibody design is the process of creating or modifying immunoglobulin sequences and structures to achieve desired antigen recognition and molecular properties. Workflows may create binders from an antigen structure, redesign complementarity-determining regions on a framework, or optimize an existing antibody. Antibody-specific numbering, paired chains, loop geometry, germline context, developability, and immunogenicity make this a distinct protein-design task.

The starting point determines the method. De novo antigen-conditioned models can propose antibodies, Fabs, or nanobodies; inverse-folding methods redesign sequences for an existing antibody structure; and CDR-focused methods preserve more of the framework while changing selected binding loops.

Predicted complexes and model scores help prioritize candidates but do not establish affinity, specificity, expression, or safety. Keep heavy–light pairing, numbering scheme, framework identity, antigen structure, and designed-region boundaries with every candidate.

When to use antibody design

• Best fit. Generating antigen-specific antibodies, Fabs, nanobodies, or redesigned CDRs
• Required starting evidence. An antigen structure, framework or antibody context, intended format, and binding assays

Benefits of antibody design

• Focused search. Uses antibody-specialized representations
• Connected evidence. Supports multiple therapeutic formats
• Testable candidates. Connects binding and developability review

Primary limitations

• Model scope. Antigen structures may omit relevant states
• Score uncertainty. Predicted affinity is not measured affinity
• Experimental requirement. Immunogenicity cannot be eliminated computationally

Antibody design methods

Antigen-conditioned generators propose antibody structures or sequences around a target surface. Framework-aware and CDR-redesign methods constrain more of the molecule, which can reduce search space while preserving known biophysical properties.

Antibody inverse folding and humanness models answer different questions. A sequence can be structurally compatible yet carry developability or immune-risk liabilities, so those assessments should remain separate.

How to run antibody design online

Use the workflow as an inspectable computational funnel. Preserve the native output of each method, apply explicit acceptance gates, and keep the evidence behind every selected and rejected candidate.

1. Prepare antigen. Prepare the antigen structure and define the intended epitope, biological assembly, and relevant glycans.
2. Choose format. Choose antibody, Fab, or nanobody format and supply a compatible framework where required.
3. Generate candidates. Generate diverse candidates while retaining chain pairing, numbering, settings, and model outputs.
4. Review structures. Inspect CDR geometry, predicted interfaces, clashes, liabilities, humanness, and developability.
5. Test binding. Express selected candidates and measure affinity, specificity, competition, and functional activity.

How to evaluate antibody design results

Review chain pairing, CDR numbering, loop conformations, epitope contacts, clashes, buried surface, sequence liabilities, humanness, aggregation risk, and agreement across structural models.

Experimental testing should include expression, monodispersity, affinity, specificity, off-target binding, competition or epitope mapping, and a function-relevant assay.

Experimental validation and handoff

Keep antigen and framework structures, numbering scheme, chain pairing, designed regions, model settings, and all candidate sequences. Confirm binding, specificity, and function experimentally.

Export structures, sequences, settings, scores, logs, and selection criteria together. A reproducible handoff makes computational assumptions visible to the team planning synthesis, expression, biophysical characterization, and functional assays.

How antibody design works

Compare RFantibody, mBER, and BoltzGen designs from antigen and framework structures.

1. Prepare antigen. Prepare the antigen structure and define the intended epitope, biological assembly, and relevant glycans.
2. Choose format. Choose antibody, Fab, or nanobody format and supply a compatible framework where required.
3. Generate candidates. Generate diverse candidates while retaining chain pairing, numbering, settings, and model outputs.
4. Review structures. Inspect CDR geometry, predicted interfaces, clashes, liabilities, humanness, and developability.
5. Test binding. Express selected candidates and measure affinity, specificity, competition, and functional activity.

Inputs and outputs

Check formats before running, then inspect and download the result from every workflow step.

Inputs

• Design input. PDB FASTA JSON TXT An antigen PDB and, when required, an antibody framework or existing variable-domain structure.

Outputs

• Design and review outputs. PDB FASTA CSV JSON Paired antibody or nanobody sequences, complex models, rankings, interface evidence, and developability results.