Antibody Evolution Service

Service Code: EG-SVC-AB-EVOL
Custom-scoped service — pricing by quotation
Request a Quote
Service Description
ServiceAntibody Evolution
OverviewA data-sovereign, AI-assisted design-build-test-learn campaign for improving an existing antibody or advancing a measured antibody library. The workflow supports paired VH/VL antibodies and VH-only formats, including VHH/nanobody starting points, while preserving exact sequence lineage, chain pairing, biological constraints, and parent controls.
Typical inputsStarting heavy/light-chain or VH-only FASTA sequences; protein or peptide antigen; optional antigen or complex structure; desired mutation scope; assay constraints; protected CDR, Vernier-zone, interface, catalytic, or epitope-contact positions; and, when available, lineage-resolved binding, enrichment, expression, stability, or specificity measurements. Antigen-only discovery can be scoped, but maturation requires a starting clone or imported candidate pool.
Campaign workflow
  1. Validate residues, chain pairing, sequence identity, provenance, objectives, and stopping rules.
  2. Assign IMGT-based numbering and define auditable CDR, framework, Vernier, interface, or explicit-position mutation masks.
  3. Generate bounded candidate panels using conservative and hydrophilic substitutions where scientifically appropriate, ESM- or AbLang-family priors, and structure-conditioned proposals such as IgDesign, AntiFold, ProteinMPNN, or RFantibody only when their inputs, runtime, weights, and use terms pass the campaign gate.
  4. Retain separate evidence for sequence liabilities, naturalness, diversity, solubility/aggregation, structure, complex plausibility, and physical or ensemble checks; unlike scores are not averaged into a universal confidence value.
  5. Lock exact parent and process controls plus a diverse ranked experimental panel and prospective assay plan.
  6. Incorporate measured results into a lineage-aware, uncertainty- and diversity-aware active-learning cycle when sufficient target-specific data exist, with group-separated calibration and out-of-domain abstention where supported.
Humanization matrixPaired VH/VL humanization preserves CDR, Vernier-zone, and VH/VL-interface constraints while comparing operational Sapiens, HuMatch, Hu-mAb-style, protected germline/template, and structural-consensus evidence as separate columns. OASis, AbNatiV, HuDiff, CUMAb, HuAbDiffusion, and related methods are added only when their implementation, reference database, model weights, transfer policy, and commercial-use terms are cleared. Agreement prioritizes variants; it does not demonstrate retained binding or stable expression.
Structure & format optionsAvailable ABodyBuilder2, NanoBodyBuilder2, IgFold, AlphaFold/ColabFold, Boltz, Chai, or related adapters can contribute method-specific structural evidence and cross-provider consensus. VHH redesign can use scaffold-preserving CDR proposals and real-sequence corpus priors. Bispecific projects can use exact topology validation for tandem VHH, DART, knob-into-hole, common-light-chain IgG, or evidence-declared CrossMab formats, with both intended target interfaces reviewed independently. Every external engine remains installation-, model-, data-, and license-gated.
DeliverablesA campaign brief and constraint contract; validated parent and candidate VH/VL or VHH FASTA; numbered sequence and mutation maps; candidate CSV and lineage records; method-separated humanization, developability, structure, and uncertainty matrices; parent and assay controls; a ranked experimental panel; recommended wet-lab assays; and a hash-bound provenance package in NextELN.
Privacy & traceabilityCore sequence analysis and controlled GPU workflows run on Excellgen-managed infrastructure. NextELN records inputs, parameters, code/model/weight/data versions, rejected candidates, measurements, decisions, and review history so the evidence remains auditable alongside every exact sequence.
Scientific boundaryThis service produces computationally ranked hypotheses. Humanization, language-model, naturalness, developability, pLDDT, PAE, pTM, ipTM, docking, and computed-energy scores do not establish affinity, specificity, expression, stability, immunogenicity, efficacy, or safety. Improvement claims require prospective SPR/BLI or equivalent binding measurements plus format-specific expression titre, SEC monomer/HMW, thermal stability, specificity, and functional assays.
EngagementExplore the scientific workspace at NextELN.com. Every campaign is custom scoped. Contact Excellgen for a technical consultation and quotation with the target, exact starting sequences, available structures and measured data, intended format, protected positions, and experimental objective.
Service Documents
DatasheetEG-SVC-AB-EVOL Product Datasheet
Download PDF
Related Services
Protein Engineering Service
Service Code: EG-SVC-PROT-ENGRequest a quote
View