Skip to course content
Learn CADD
Applied extension

Module 16: Protein, Antibody & Protein-Protein Modeling

Extend structure-based design from small-molecule pockets to flexible protein interfaces, antibody paratopes, multispecific architectures, and developability-aware protein engineering.

Learning outcomes

  • Explain how protein and antibody modeling differs from small-molecule docking.
  • Build an ensemble-aware antibody or engineered-protein model.
  • Use experimental restraints to constrain and validate protein-protein docking.
  • Evaluate interface quality alongside stability, aggregation, and other developability risks.

Interactive interface-ranking exercise

Which protein-complex pose would you advance?

Raw energy favors Pose A. Add independent evidence to test whether that ranking survives clashes, experimental restraints, conformational uncertainty, and developability risk.

Current decision

Advance Pose A

96/100

This decision uses only the docking-energy rank and is a hypothesis, not a validated interface.

#1 Pose A

consensus 96
Energy
96
Restraints
42%
Clashes
8
Ensemble
1/5
Developability
35

#2 Pose B

consensus 82
Energy
82
Restraints
91%
Clashes
1
Ensemble
4/5
Developability
72

#3 Pose C

consensus 70
Energy
70
Restraints
76%
Clashes
0
Ensemble
3/5
Developability
90

Teaching model: these normalized values illustrate evidence integration; they are not physical energies or a universal scoring function. In a real project, predefine acceptance criteria and challenge the leading poses experimentally.

1. Why protein therapeutics need a different modeling stack

Antibodies, nanobodies, engineered binders, enzymes, and multispecific proteins operate through large, flexible interfaces. Their models must represent sequence, fold, loop uncertainty, oligomerization, glycosylation, electrostatics, conformational change, and manufacturability.

Small molecule

A compact ligand is usually docked into a comparatively localized pocket. Search emphasizes ligand conformation, pose, and receptor flexibility.

Protein-protein complex

Two large surfaces must be oriented while side chains, loops, and sometimes domains reorganize. Shape complementarity alone produces many false poses.

Antibody-antigen complex

Six CDR loops create the paratope, but framework residues, orientation of variable domains, glycans, and long CDR-H3 conformations can influence recognition.

Multispecific format

Multiple binding arms introduce valency, linker geometry, avidity, competing target sinks, assembly risk, and tissue-distribution constraints.

2. Antibody structure and sequence annotation

ElementModeling questionCommon failure mode
FrameworkIs the template close in sequence and canonical geometry?A poor framework shifts the relative orientation of binding loops.
CDR loopsWhich numbering scheme and boundary definition are being used?Residue positions are compared across incompatible schemes.
CDR-H3How uncertain are length, kink, base, and loop conformations?One highly uncertain model is treated as a solved paratope.
VH/VL orientationDoes the template support the intended interface geometry?Correct local loops are combined with the wrong domain orientation.
Fc and glycansAre effector function, FcRn interaction, and glycosylation relevant?A truncated model is used for whole-antibody conclusions.

3. Building an antibody or engineered-protein model

  1. 1

    Curate the sequence

    Confirm chain boundaries, signal peptides, mutations, disulfides, numbering, construct tags, linkers, and intended oligomeric state.

  2. 2

    Select templates or predictions

    Choose frameworks and domain orientations using sequence and structural compatibility. Use prediction confidence to define, not hide, uncertain regions.

  3. 3

    Model loops and side chains as an ensemble

    Generate alternatives for CDR-H3, engineered loops, and interface side chains. Filter clashes and poor stereochemistry before complex modeling.

  4. 4

    Add relevant chemistry

    Represent disulfides, glycans, protonation, post-translational modifications, metals, and linker geometry when they affect the question.

  5. 5

    Validate against data

    Use known mutagenesis, epitope mapping, competition, crosslinking, HDX, cryo-EM density, or homologous complexes to challenge the model.

4. Protein-protein and antibody-antigen docking

Protein docking explores rigid-body orientation first, then refines interfaces. Unconstrained global docking is difficult because the surface is large and flexibility is expensive; experimental restraints sharply reduce the search space.

EvidenceHow it constrains dockingCaution
Known epitope/paratope residuesDefines attractive or ambiguous interaction restraintsA functional residue may act indirectly rather than contact the partner.
CrosslinksRestricts pairs to a distance rangeAccount for linker length, side-chain geometry, and uncertainty.
MutagenesisPrioritizes interface patches and tests refined posesLoss of binding can result from destabilization.
Competition or homologous complexRestricts the face and orientation of bindingHomologous partners may use different loops or angles.
Density or low-resolution shapeFilters rigid-body poses globallyFlexible regions may not be resolved.

5. Interface scoring and validation

Geometry

Check buried surface area, shape complementarity, clashes, cavities, unsatisfied polar groups, and interface planarity. Remove impossible poses before interpreting energy.

Chemistry

Inspect salt bridges, hydrogen-bond networks, aromatic and cation-pi contacts, hydrophobic patches, water mediation, and electrostatic complementarity.

Robustness

Compare scoring functions, refinement protocols, starting structures, and ensemble members. Stable conclusions survive reasonable modeling choices.

Experimental falsification

Select mutations or binding measurements that distinguish competing poses. The most useful model proposes a test that could prove it wrong.

6. Protein engineering and developability

RiskSequence/structure signalsEngineering response
Low stabilityBuried unsatisfied groups, cavities, poor packing, exposed hydrophobicsStabilize the core or interface while preserving function.
AggregationLarge hydrophobic or charged surface patches and flexible exposed segmentsReduce patchiness, improve colloidal behavior, and test concentration dependence.
PolyspecificityBroad hydrophobic or electrostatic complementarityBalance paratope chemistry and screen nonspecific binding experimentally.
Chemical liabilitiesDeamidation, isomerization, oxidation, cleavage, or unpaired cysteine motifsRemove hotspots when compatible with activity and structure.
ImmunogenicityPotential T-cell epitopes, non-human framework content, aggregatesHumanize and de-risk with sequence, structure, and experimental evidence.
Poor multispecific geometrySteric occlusion, strained linkers, incompatible arm spacingModel full-format architecture and test simultaneous engagement.

Knowledge check

Self-Assessment ChallengeQuestion 1 of 4

Why is protein-protein docking generally harder than docking a small molecule?