HarjasG/egfr-domain3-anchor-map

EGFR Domain III Titratable Anchor Map

Residue-level annotation of the human EGFR domain III epitope face, built for designing pH-conditional binders — ones that engage at pH 6.5 and release at pH 7.4.

Derived from PDB 6ARU chain A (EGFR ectodomain, cetuximab-Fab-bound, 3.2 Å). Numbering is UniProt P00533-1 precursor numbering throughout.

Why this exists

Designing an acidic-ON binder means engineering titratable contacts that gain favourable interaction when protonated. The linkage ceiling is 10^ΔpH ≈ 8× per site over 6.5 → 7.4, so clearing a "no detectable binding" bar needs three titratable pairs minimum. Finding three that a single binder can span simultaneously turns out to be the binding constraint, and nothing published maps them.

This dataset is that map.

Contents

residues.csv — every domain III residue

column	meaning
precursor_num	UniProt P00533-1 numbering
resname	three-letter code
rel_sasa	relative solvent accessibility (Shrake-Rupley / Tien et al. 2013 max-ASA)
is_epitope	within 5 Å of the bound cetuximab Fab
min_partner_dist	min heavy-atom distance to Fab, Å
conserved_mouse	identical in mouse EGFR (UniProt Q01279)
pka	PROPKA3 predicted pKa

anchors.csv — surface titratable groups eligible as design anchors

class	target residue	complementary binder residue	switch sits on
A	Asp / Glu	His	binder
B	His	Asp / Glu	target

Class B is the better anchor where available: the switching residue is on EGFR, so it is present in both assay conditions by construction, and it spares the binder from carrying four histidines (desolvation cost, aggregation risk, pI drift). Domain III carries H370 and H433, the residues the cross-reactive pH-dependent antibody 14C07/G5V2 exploits.

patches.json — anchor sets within mutual spanning distance, ranked. Flags meets_spec_v2 for patches with ≥3 anchors, ≥1 Class B, all mouse-conserved. Includes side-chain tip coordinates and outward vectors for placing the complementary binder residue.

Headline numbers

Residues annotated	609
On the cetuximab contact face	26
Identity to mouse EGFR	88.8% (published: 88%)
Surface titratable anchors in domain III	19 (16 Class A, 3 Class B)
Patches meeting ≥3 pairs, ≥1 Class B, all conserved	4

HIS433 — pKa 6.22, relative SASA 0.648, identical in mouse, inside the cetuximab contact footprint. ~37% protonated at pH 6.5 against ~6% at 7.4. The best available anchor on the protein, and the only one servicing all three design objectives at once.

HIS358 — pKa 6.26, conserved, but outside the Fab footprint.

HIS383 — excluded. Arginine in mouse, so titratability is lost entirely.

How it was built

epitope_prep.py — Biopython for structure handling and SASA, PROPKA3 for pKa, global pairwise alignment against the mouse ortholog for conservation. Epitope face derived from the bound Fab rather than a literature residue list.

Using cetuximab to define the epitope is permitted under the Adaptyv competition's de novo rule. No Fab coordinates or sequence enter any design.

Caveats

• pKa values are PROPKA3 predictions on a single crystal conformation. Treat as a tier-1 screen, not ground truth — a static structure cannot see a salt bridge that dissociates in 5 ns of MD.
• Conservation comes from sequence alignment, not a structural superposition of the mouse ortholog.
• rel_sasa is computed on the isolated chain, so residues buried by domains II/IV in the tethered conformation will read as more exposed than they are in the full ectodomain.

Citation

If this is useful, cite the structure (Sickmier et al., PDB 6ARU) and PROPKA3 (Olsson et al., JCTC 2011).