Improved protein binder design using β-pairing targeted RFdiffusion

Area of Science:

• Protein engineering
• Computational biology
• Structural biology

Background:

• Designing high-affinity protein binders for hydrophilic targets is difficult.
• Existing computational methods have limitations in targeting specific protein surfaces.

Purpose of the Study:

• To develop and validate a conditioned RFdiffusion approach for designing protein binders.
• To generate binders targeting edge-strand sites on various proteins, including KIT, PDGFRɑ, ALK-2, ALK-3, FCRL5, NRP1, and α-CTX.

Main Methods:

• Utilized conditioned RFdiffusion to generate protein scaffolds with geometrically matched extended β-sheets.
• Complemented polar groups on target proteins with hydrogen bonding groups on designed binders.
• Tested designed binders against multiple protein targets.

Main Results:

• Achieved higher binding affinities (pM to mid nM) and success rates compared to unconditioned RFdiffusion.
• Demonstrated high specificity of designed binders due to precise geometric customization and additional interactions.
• Validated the design approach with a co-crystal structure of a binder-KIT complex, showing high structural accuracy.

Conclusions:

• Conditioned RFdiffusion enables robust generation of binders for hydrophilic, exposed β-strand surfaces.
• This method significantly expands the scope of computational protein binder design.
• The precise customization of binder-target interactions leads to high affinity and specificity.