Target-conditioned diffusion generates potent TNFR superfamily antagonists and agonists

Area of Science:

• Computational biology
• Protein engineering
• Structural biology

Background:

• Designing high-affinity protein binders for challenging targets, particularly those with flat and polar surfaces like tumor necrosis factor receptor 1 (TNFR1), remains a significant hurdle in current protein design methodologies.
• Existing methods often fall short in achieving precise shape complementarity compared to native protein complexes.

Purpose of the Study:

• To investigate the efficacy of a novel computational approach utilizing free diffusion from random noise for generating shape-matched protein binders.
• To test this approach on the challenging target TNFR1 and explore its potential for designing binders with tunable affinity and specificity.

Main Methods:

• A computational design strategy based on free diffusion from random noise was employed.
• The method was applied to design binders for tumor necrosis factor receptor 1 (TNFR1).
• Partial diffusion techniques were used to modulate the specificity of the designed binders.

Main Results:

• The study successfully generated protein designs with low picomolar affinity for TNFR1.
• Binder specificity could be precisely switched to other related protein family members through partial diffusion.
• Designed proteins functioned as antagonists or superagonists for OX40 and 4-1BB targets when presented at higher valency.

Conclusions:

• Free diffusion from random noise is a viable computational strategy for designing high-affinity and high-specificity protein binders, even for difficult targets.
• This in silico approach offers precise control over binder function (antagonist/agonist) and specificity, applicable to pharmacologically relevant targets.
• Computational protein design is poised to replace traditional methods like immunization and random screening for creating therapeutic proteins.