AF3Design: Selectivity–Aware Nanobody Design with AlphaFold3

Designing antibodies that discriminate between highly similar targets represents a fundamental challenge in computational protein engineering, particularly for T cell receptor-like (TCR-like) nanobodies targeting peptide-MHC complexes. We present \textit{AF3Design}, the first computational framework for nanobody design that integrates binding selectivity directly into the optimization process rather than relying on post-hoc filtering. Our approach leverages AlphaFold3's structure prediction capabilities through a gradient-free genetic algorithm, optimizing a composite fitness function with explicit off-target penalties. These terms include contact-based rewards for discriminating residues and ${\Delta\Delta}$G-based differential binding predictions. We validate our method on the challenging task of designing nanobodies against peptide-MHC complexes that differ by as little as a single residue. AF3Design significantly improves peptide-facing interface confidence over baselines and enriches binders that preferentially contact the discriminating peptide residue(s), with further $\Delta\Delta$G gains after staged activation. Our results show that optimizing selectivity during search yields highly discriminative candidates suitable for precise molecular recognition.