Choose the Site Before the Binder

Two binders can recognize the same receptor and do different jobs. One may block a partner; another may leave that interaction intact. A positive binding assay tells a team that a molecule engages its target. It does not, by itself, tell them whether it engages the useful surface.

At Crane, protein engineers turn the intended mechanism into a spatial brief: the receptor feature to block or engage, the surfaces to avoid, the orientation a payload may need, and the measurements that would prove the hypothesis.

The measured cryo-EM structure PDB 8Q6J captures one HER2 ectodomain with both trastuzumab and pertuzumab Fabs. It makes the strategic contrast visible on the same receptor.

Trastuzumab occupies the membrane-proximal domain IV epitope also measured in PDB 1N8Z. Pertuzumab binds near the center of domain II in PDB 1S78, beside a receptor dimerization surface.

The pertuzumab structure and mutagenesis study supports a steric model for interfering with receptor dimerization. The structure does not establish that any new binder at that site will reproduce the same affinity, signaling effect, or cellular response.

Once the intended receptor event is clear, an epitope can become a design constraint. Hotspots can define where contacts should form. Negative constraints can keep a candidate away from glycans, conserved family surfaces, or geometry that would obstruct the intended format.

A published HER2 domain IV binder program used iterative computational design and experimental feedback to refine a site-specific LRR binder. That loop matters. It is not a one-shot demonstration that choosing a patch and running a model produces a development candidate.

The deliverable is a spatial brief that explains why this surface is worth pursuing. It names the accessibility and orientation constraints and the result that would make the team reject the proposed mechanism. A predicted interface then carries a testable argument into the synthesis decision.

Our PD-L1 cyclic-peptide study shows how we evaluated a different interface problem computationally. It reports design and structural assessment, not measured binding or a validated therapeutic outcome. It is an example of Crane's execution alongside the published HER2 evidence discussed here.