# Executive Summary — EAC Dual-Arm Discovery-to-Design Preprint

**Paper:** *Steering into competitive whitespace: a public-data, computation-driven campaign nominates GUCY2C and DKK1 and delivers de novo binder leads across the esophageal adenocarcinoma trajectory.*
**Status:** bioRxiv-style preprint, acceptance-ready after two synthetic peer-review rounds (Major → Minor revision; all essential revisions resolved or honestly scoped). **All results are computational (in-silico); no experimental validation has been performed.**

## What the paper does
It converts a completed EAC drug-discovery campaign into a submission-ready preprint. Working entirely from public data — TCGA/cBioPortal genomics, Open Targets, ClinicalTrials.gov, Drugs@FDA, GTEx, and AlphaFold — it (1) confirms EAC as a copy-number-driven cancer whose druggable drivers are clinically crowded, (2) maps competitive whitespace, (3) applies an explicit, disclosed druggability triage to nominate two leads spanning the disease trajectory, and (4) designs de novo protein binders for both.

## The two leads
- **GUCY2C** (treatment arm) — a GI-restricted surface antigen for an anti-GUCY2C × anti-CD3 T-cell engager in advanced EAC. GTEx confirms 85% GI-restricted normal expression (top non-GI tissue ~1.7 TPM). Top binder gucy2c_bb2: ipTM 0.915, but flagged with a high-alanine developability liability.
- **DKK1** (interception arm) — a secreted Wnt modulator for a neutralizing trap in high-risk Barrett's, with DKN-01 clinical precedent (in advanced disease, not the precursor setting). Top binder dkk1_bb1: ipTM 0.858 and the cleanest design on both interface confidence and sequence composition — the most robust lead in the set.

## Key numbers (all traceable to artifacts/public sources)
- EAC genomics: TP53 87%, CDKN2A del 39%, CCND1 amp 35% (TCGA, n=182).
- Triage: 16 candidates; disclosed additive rubric reproduces all scores to 0.005; GUCY2C ranks #1 among biologic-addressable treatment targets in 80.7% of 20,000 weight-perturbation draws — but this is contingent on its optimistic selectivity score: under the conservative recalibration (O-2 unresolved), CDH17 overtakes it, so O-2 decides lead choice as well as safety and CDH17 is carried as backup. DKK1 is the only addressable interception target after the filter.
- Design: 15/16 designs pass ipTM>0.5 (pilot-scale, n=8 backbones/target, single seed).
- Interface residues map to UniProt P25092 153–402 (GUCY2C ECD) and O94907 187–208 (DKK1 CRD2).

## Why it matters
It is a reproducible, honesty-gated template for early differentiated target-and-lead generation: it routes around the saturated HER2/PD-1/VEGF field, attacks EAC at two under-exploited trajectory positions (including molecular interception, where nothing is approved), and delivers concrete designed binders a wet-lab program could take forward.

## Honest verdict and what remains
The framing matches the evidence: these are pilot-scale in-silico leads, not validated therapeutics. Two make-or-break questions are downstream and unresolved — the **GUCY2C normal-GI therapeutic window** for a T-cell engager (O-2) and the **DKK1 dysplasia-regression surrogate endpoint** for a prevention trial (O-4). The decisive near-term computational step (orthogonal AF-Multimer/Chai-1 fold-back plus specificity controls, O-1) is named, not claimed. One follow-up query — GUCY2C tumor-cell mRNA in EAC — could not be run through available tooling and is flagged as the top priority rather than fabricated.