mltox
Request a demo

Protein and antibody engineering

mltox Proteins & Biologics Engineering

Map sequence-based immunogenicity and developability hypotheses, then propose conservative edits for expert and experimental review.

Discuss a biologics pilot View a sequence example
Protein and antibody candidates move through sequence, structure, developability, conservative engineering, and focused assay review.

The decision

A sequence liability is easiest to address while design freedom remains

Immunogenicity, aggregation propensity, sequence instability, intrinsic solubility, and structural exposure can create competing design hypotheses. Function, expression, measured biophysics, and clinical ADA remain experimental questions.

  1. Immunogenicity screening

    Map predicted HLA class I and II presentation signals, compare hotspots, and retain the relevant biological limits.

  2. Developability review

    Inspect aggregation, oxidation, deamidation, isomerization, glycosylation, composition, and solubility hypotheses.

  3. Structure-aware refinement

    Use predicted structure and confidence to re-rank exposed, buried, or uncertain regions.

  4. Constrained sequence hypotheses

    Review conservative substitutions predicted to reduce an epitope, then test function, binding, expression, and biophysics in the lab.

BIO-01 | Immunogenicity screening

Compare predicted presentation risk at the residue and candidate level

Use class-II epitope load and humanness as the primary ADA-related sequence signal, with a supplementary class-I panel. The output prioritizes review; it does not predict a clinical anti-drug-antibody rate.

  1. CompareRank candidates by load and humanness
  2. InspectLocate promiscuous residue-level hotspots
  3. TestFocus ex vivo or bioanalytical follow-up
  4. AbstainStop when modeled context is insufficient
Current MLTox biologics portal showing the seven-allele HLA-DR panel, non-human epitope load, humanness, hotspot positions, and candidate sequences
Current portal immunogenicity module. Candidate-level context and residue-level evidence now occupy one aligned review surface. This estimates presentation risk, not a clinical ADA rate.

BIO-02 | Developability

Treat developability as a profile, not a verdict

Review deamidation, isomerization, oxidation, glycosylation, cysteine, cleavage, fragmentation, and CamSol-style intrinsic-solubility flags together. They are sequence hypotheses, not measured manufacturability, formulation, or stability.

Prototype visual: A linked sequence and structure view where motif flags are qualified by exposure and confidence before routing to different aggregation, stability, solubility, or degradation assays.
A linked sequence and structure view where motif flags are qualified by exposure and confidence before routing to different aggregation, stability, solubility, or degradation assays. Prototype artwork with illustrative data.

BIO-04 | Structure-aware refinement

Use structure only when it changes the priority

When GPU folding is enabled, predicted solvent exposure re-weights both axes and supports a 3D review. Long chains may be split at linker-like regions; segment placement relative to one another is not predicted.

Portal interaction, public example. Drag to rotate and scroll to zoom. This embedded viewer uses the same self-hosted 3Dmol.js approach as the MLTox portal. The displayed structure is the public experimental HER2–trastuzumab Fab complex PDB 1N8Z; submitted MLTox sequences use predicted ESMFold structures and must be interpreted with pLDDT and exposure limits.

Interactive portal viewer HER2–trastuzumab Fab complex PDB 1N8Z · experimental example
Loading structure…
Fab light chain Fab heavy chain HER2

BIO-03 | Constrained sequence design

Generate fewer, better-defined hypotheses for the lab

For a non-human epitope, propose conservative BLOSUM62 substitutions while fixing Cys, Pro, and glycosylation motifs. The model predicts epitope reduction only; every candidate requires functional and experimental validation.

Prototype visual: A before-and-after sequence review showing an original hotspot, protected regions, proposed substitutions, aligned liability changes, tradeoffs, neoepitope checks, and a test plan.
Before-And-After Review A before-and-after sequence review showing an original hotspot, protected regions, proposed substitutions, aligned liability changes, tradeoffs, neoepitope checks, and a test plan. Prototype artwork with illustrative data.

Know what the biologics engine does not establish

The most useful result is a focused hypothesis with a clear experimental handoff, not an apparently complete verdict.

Epitope and humanness signals

Prioritize sequence regions and constructs for review.

Not modeled

Clinical ADA incidence

Patient-level and product-level clinical ADA rates still require relevant clinical and experimental evidence.

Substitution hypotheses

Compare predicted liability changes before testing.

Not guaranteed

Function preservation

Binding, activity, folding, expression, and stability must be checked after a proposed sequence change.

Intrinsic sequence profile

Expose regions that may merit developability review.

Not measured

Formulation behavior

Solubility and aggregation under real concentration, buffer, process, and storage conditions remain experimental.

Sequence-derived evidence

Focus the next engineering or assay question.

Outside scope

CRS, polyreactivity, and off-target biology

Sequence alone does not establish agonism, cytokine-release risk, target biology, or antibody polyreactivity.

Team workflow

One record for the team, separate judgments for each discipline

Protein and antibody engineering

Compare constructs

Review candidate tradeoffs while protecting functional regions.

Immunogenicity

Challenge immune assumptions

Review HLA panels, hotspots, immune context, and assay strategy.

Developability and formulation

Connect to measurement

Carry sequence hypotheses into relevant biophysical and early CMC work.

Structural biology and design

Inspect structural limits

Review confidence, exposure, interfaces, and what predicted structure omits.

Focused evaluation

Evaluate a construct family and the decisions around it

Use representative sequences with known results or planned experiments. Define the HLA panel, liabilities, protected regions, structural comparators, assays, and success measures before analysis.

Discuss a biologics pilot

Judge the workflow on

  • Candidate prioritization and useful tradeoffs.
  • Hotspot actionability and protected constraints.
  • Appropriate caveats and abstention.
  • Experimental focus and handoff quality.
  • A required exit: adopt, restrict, revise, or stop.