Translate a defined antigen region or functional binding objective into a prioritized panel of novel antibody candidates through integrated framework selection, CDR generation, complex modeling, interface assessment, and developability-aware ranking.
Request a Design Assessment →Conventional antibody discovery can produce strong binders without necessarily reaching the epitope that drives the desired mechanism. An epitope-guided design project begins with a different question: which antigen region must be engaged, avoided, or sterically blocked to support the intended biological outcome? CD ComputaBio converts that question into explicit structural and sequence constraints, then explores antibody candidates capable of satisfying them.
The service is designed for projects that have an antigen sequence and a defensible structural model, and is strongest when residue-level epitope information, competition data, mutational evidence, or a known ligand–receptor interface is available. When the epitope is uncertain, we can first assess candidate surface regions and document the uncertainty rather than treating a predicted site as experimentally confirmed.
Computational design narrows an otherwise enormous search space. It does not establish binding, affinity, specificity, expression, or biological activity. The output is a traceable candidate panel and a validation strategy that helps teams allocate synthesis and screening capacity more rationally.
The project is configured around the available structural evidence, desired antibody format, experimental capacity, and novelty objective. "De novo" may involve designing a new paratope on a selected framework or exploring broader framework–CDR combinations; the scope is stated explicitly in the project plan.
Design selected CDRs, including CDR-H3 where appropriate, on one or more structurally compatible frameworks. This route provides tighter control of sequence context and is suitable when format, framework family, or human-likeness requirements are already defined.
Generate paratope concepts around a specified antigen patch, considering surface geometry, accessibility, electrostatics, hydrogen-bond opportunities, hydrophobic features, and the approach orientations available to the selected antibody format.
Retain multiple sequence families, CDR solutions, or binding orientations instead of returning only one highest-scoring model. Diversity-aware selection reduces dependence on a single scoring assumption and supports more informative experimental testing.
Translate computationally favored positions and tolerated residue classes into a controlled library architecture. Suggested diversity is matched to the intended display or screening capacity, with liabilities and structurally sensitive positions excluded where possible.
Candidate generation is coupled to antigen assessment, complex modeling, interface analysis, and risk filtering. Each method is selected because it answers a project question; no single score is treated as proof of experimental performance.
We review sequence identity, structural coverage, missing loops, alternate conformations, oligomeric state, glycosylation, membrane context, and the evidence supporting the proposed epitope. Surface accessibility, conservation, charge, hydrophobicity, and potential steric barriers are assessed before sequence generation begins.
Candidate frameworks and variable-domain orientations are evaluated for CDR reach, interface geometry, format compatibility, and clashes with the antigen. A modeled pose is treated according to its confidence; uncertain docking solutions may be retained as competing hypotheses rather than collapsed into a false consensus.
Sequence and backbone alternatives are sampled under antibody-specific structural constraints. CDR length, loop geometry, residue composition, framework contacts, and heavy–light chain pairing are considered together. The design space can be restricted by germline preference, human-likeness, excluded motifs, or client-defined residues.
Candidate complexes are refined and compared using contact recovery, buried surface area, geometric complementarity, polar networks, electrostatic compatibility, unsatisfied groups, clashes, and energetic indicators. Molecular dynamics or higher-cost free-energy calculations may be used for a smaller shortlist when justified.
Designs are screened for sequence liabilities such as problematic hydrophobic patches, extreme charge distributions, aggregation-prone motifs, chemical modification hotspots, unusual cysteine patterns, and framework–CDR incompatibilities. These predictions are risk flags, not substitutes for expression and biophysical assays.
Final selection balances predicted epitope engagement, structure confidence, interface quality, sequence plausibility, developability indicators, and diversity. Component scores and exclusion reasons are retained so that project teams can review why each candidate was advanced, held as an alternative, or removed.
| Stage | Key activities | Decision output |
|---|---|---|
| 1. Objective and acceptance criteria | Define antigen, epitope evidence, intended mechanism, antibody format, species profile, novelty constraints, experimental capacity, and desired readouts. | A design brief distinguishing required properties, preferred properties, and assumptions that need testing. |
| 2. Input and structure audit | Check antigen sequence, construct boundaries, structure quality, chain assignment, glycosylation, biological assembly, missing regions, and compatibility of experimental or predicted models. | A data-readiness assessment, structural confidence map, and list of gaps that may limit interpretation. |
| 3. Epitope and orientation definition | Characterize the target surface, establish permitted approach regions, evaluate framework/CDR reach, and generate alternative orientations when the binding pose is uncertain. | One or more explicit epitope–orientation hypotheses with traceable constraints. |
| 4. Candidate generation | Generate CDR and paratope variants under structural, sequence, framework, human-likeness, and client-defined constraints; retain independent sequence families. | A diverse design pool with design lineage and constraint compliance records. |
| 5. Modeling and multi-parameter ranking | Build candidate complexes, refine interfaces, evaluate interactions and structural confidence, flag developability risks, cluster similar candidates, and test ranking sensitivity. | Ranked candidates grouped as binding-focused, balanced, conservative, and structurally diverse options. |
| 6. Handoff and validation planning | Select an experiment-sized panel, define controls, recommend expression and binding assays, and document which computational hypotheses each experiment will test. | A synthesis-ready candidate package, methods report, limitations, and staged validation plan. |
Published antibody-design frameworks demonstrate why CDR sequence cannot be evaluated independently of loop geometry, framework support, antigen orientation, and local structural relaxation. Iterative sampling can explore these coupled variables, but performance remains conditional on the quality of the starting antigen model, the realism of the candidate binding pose, the suitability of the scoring functions, and the similarity of the project to data represented in the underlying methods.
For that reason, CD ComputaBio uses ensemble and evidence-aware decision logic. Experimental structures are distinguished from predicted structures; high-confidence regions are separated from uncertain loops; competing poses may be carried forward; and candidate ranks can be stress-tested against alternative weighting schemes. The aim is not to produce a visually convincing complex, but to identify candidates whose rationale remains credible under several reasonable modeling choices.
Experimental confirmation is essential. Small-scale expression and purity assessment should precede interpretation of negative binding results, while orthogonal binding methods help distinguish assay artifacts from genuine recognition. Affinity and kinetic measurements, competition or epitope-confirmation experiments, and mechanism-relevant functional assays should be staged according to the original design objective.
Not every item is mandatory, but uncertainty in the inputs changes the appropriate design depth and the confidence of candidate prioritization.
Agreed biological objective, format, target surface, acceptance criteria, input provenance, structural confidence, excluded regions, and unresolved questions.
Prepared antigen model, target-surface map, relevant conformations, model-quality observations, and machine-readable structure files appropriate to the agreed scope.
VH/VL or single-domain candidates with CDR annotation, framework information, sequence-family assignments, design lineage, and candidate-level rationale.
Representative antibody–antigen models, interface contacts, approach orientations, interaction summaries, confidence notes, and alternative pose hypotheses where relevant.
Binding-related indicators, structural plausibility, sequence naturalness, human-likeness, predicted liabilities, diversity, and transparent advancement or exclusion reasons.
Recommended candidate panel, expression and screening priorities, controls, assay sequence, decision thresholds, and options for iterative redesign after data return.
A practical experimental panel includes candidates that test different design hypotheses. This makes negative results informative and reduces the chance that one imperfect computational metric determines the whole program.
Candidates emphasizing target-epitope contacts, geometric complementarity, and predicted interface quality. These may carry higher sequence or developability uncertainty and should be interpreted accordingly.
Candidates selected for a compromise among interface quality, sequence plausibility, structural confidence, human-likeness, and lower predicted liability. These are often suitable as the core testing set.
Candidates representing alternative CDR solutions, sequence families, or approach orientations. They help test whether the biological surface can be engaged through more than one plausible recognition mode.
Design candidates around a ligand-binding surface, receptor interaction site, protease-sensitive region, or conformationally important patch when biological activity depends on where the antibody binds rather than binding alone.
Evaluate conserved residues and structural variation across homologs or species to develop candidates intended to recognize a shared surface, while explicitly flagging sequence or conformational differences that may limit cross-reactivity.
Focus design on differentiating residues or local surface features when a program needs to distinguish related proteins, variants, or isoforms. Computational selectivity remains a hypothesis requiring counter-screening against relevant off-target antigens.
Explore structurally defined surfaces that may be underrepresented in conventional immune repertoires. Accessibility, flexibility, glycan shielding, and membrane proximity are assessed before a difficult region is treated as designable.
No, but it materially improves the definition of the design problem. If only an antigen structure and functional region are available, the project can evaluate candidate surface patches first. Predicted regions will be labeled as computational hypotheses and should not be described as experimentally mapped epitopes.
Yes, when the relevant surface is modeled with adequate confidence. We review domain boundaries, missing regions, conformational state, oligomeric context, and local uncertainty. Low-confidence epitope geometry may require several structural hypotheses or may limit the value of fine-grained interface optimization.
Both scopes can be considered. Many projects use selected frameworks and design the paratope or chosen CDRs, while broader projects explore multiple framework–CDR combinations. The deliverable will state which residues and structural components were generated, retained, or optimized.
Projects may involve conventional VH/VL variable regions, Fab or scFv concepts, VHH or other single-domain antibody formats, and defined binding domains for multispecific architectures. Format-specific geometry and downstream engineering constraints must be included during scoping.
The answer depends on design novelty, structural confidence, ranking separation, candidate diversity, assay throughput, and budget. We recommend an information-rich panel spanning several design hypotheses instead of claiming that one top-ranked sequence is sufficient.
No. Sequence generation, docking, structure prediction, and energetic scoring have model and data limitations. Experimental expression, binding, specificity, affinity, epitope confirmation, and functional testing are required before any performance conclusion can be made.
Yes. Expression, binding, kinetic, competition, mutagenesis, and functional results can be mapped back to the original hypotheses. A second round can then preserve successful sequence features, remove unsupported assumptions, refine the pose, and concentrate diversity around experimentally informative positions.
Share the antigen, structural evidence, target epitope, antibody format, constraints, and planned assays. Our team will assess feasibility and propose a fit-for-purpose computational design scope.
Submit your project details below, and our team will respond within 24 hours.
Talk to our technical team about your project!
I Want To Talk
