Viral escape-inspired framework for structure-guided dual bait protein biosensor design.

Area of Science:

• Computational biology
• Protein engineering
• Immunology

Background:

• The SARS-CoV-2 spike protein's receptor binding domain (RBD) evolves to evade antibodies while maintaining ACE2 receptor binding.
• Existing viral escape predictors often require vast data and lack generalizability for diverse protein design tasks.

Purpose of the Study:

• Introduce CTRL-V, a generalizable computational platform for designing selective binding (dual bait) biosensor proteins.
• Utilize SARS-CoV-2 RBD as a model to demonstrate CTRL-V's capability in identifying immune-evasive mutations.

Main Methods:

• Iterative design cycles combining integer optimization, stochastic sampling (PyRosetta), and deep learning (ProteinMPNN).
• Leveraging publicly available viral escape data as anchors for a generalized computational workflow.
• Application to identify mutations in Raf kinase for selective binding to Ras and Rap1a GTP.

Main Results:

• CTRL-V identified 20% of reported SARS-CoV-2 mutations linked to immune escape from a specific antibody.
• Accurately pinpointed ~70% of single point mutations in the KP.2 variant, providing structural insights into immune escape mechanisms.
• Demonstrated generalizability by identifying key mutational sites in Raf kinase for selective GTP binding.

Conclusions:

• CTRL-V offers a physics-informed, generalizable approach to biosensor design, overcoming limitations of purely data-driven methods.
• The platform successfully models viral evolution for immune escape and can be adapted for other protein engineering challenges.
• CTRL-V advances the design of selective biosensors with potential applications beyond virology, including kinase-GTP interactions.