Crowdsourced Protein Design: Lessons From the Adaptyv EGFR Binder Competition
Problem
Designing proteins that bind EGFR, a key cancer drug target, is difficult and lacks standardized benchmarks.
Why it matters
EGFR drives cell growth and cancer, so binders could enable new diagnostics and therapies.
What the AI did
Competition participants used computational and ML methods of their choice to design EGFR-binding proteins, which were then experimentally tested in Adaptyv's automated BLI pipeline.
Results
601 of 1,800+ designs characterized; optimized binder K_D = 1.21 nM; best de novo binder K_D = 82 nM
Impact
Shows crowdsourced ML protein design can yield experimentally validated EGFR binders, though benchmarks remain lacking.
Record
| Venue | bioRxiv (Cold Spring Harbor Laboratory) |
|---|---|
| Published | 2025 |
| Work type | preprint |
| Authors | Tudor‐Stefan Cotet, Igor Krawczuk, Filippo Stocco, Noelia Ferruz, Anthony Gitter, Yoichi Kurumida, Lucas de Almeida Machado, Francesco Paesani et al. |
| Citations | 26 (OpenAlex, at import time) |
| References | 79 |
| Model or method | not stated in source |
| Organization | Adaptyv |
| Primary source | https://doi.org/10.1101/2025.04.17.648362 |
| Open access | green |
| Keywords | protein design, EGFR, crowdsourcing, machine learning, binder competition |
Importance rating
significant — thousands to millions affected
Abstract calls EGFR a key drug target involved in cell growth, differentiation, and cancer development; designed binders could support therapy/diagnostics, though no scale of patients affected is given.