The 48-Hour Cancer Binder (ludocomito.dev)

A team participating in a protein design hackathon in Zurich successfully developed a binding protein targeting FGFR2, a receptor involved in certain cancers. Their objective was to create an inhibitor that selectively binds to FGFR2 without affecting the similar FGFR1 receptor, a critical requirement to avoid undesirable side effects. The design process involved identifying key regions of FGFR2, particularly the D3 domain, and utilizing ML tools such as BoltzGen to generate and rank potential protein binders based on their predicted binding affinities. This accomplishment is significant for the AI/ML community as it showcases the integration of machine learning models with biological knowledge to address real-world medical challenges. The team's binder achieved an impressive iPSAE score of 0.7 for FGFR2 while maintaining a low score of 0.1 for FGFR1, indicating high specificity. Their experience also highlights the importance of interdisciplinary collaboration and the use of various computational models to assess properties beyond binding affinity, including immunogenicity and stability. While the team did not win the hackathon, the project exemplifies the potential of ML in drug design and emphasizes the necessity of effective communication in presenting research outcomes.