Developability-Guided Antibody Design Using FlashTAP

Antibody design usually treats developability as a filter rather than an optimisation objective. This means designs which may otherwise have ideal binding properties might still be rejected due to developability issues. We integrate FlashTAP, a differentiable predictor for four classical developability metrics, at three inference-time intervention points: post-hoc repair, structure-guided inverse folding and de novo backbone design. On a fixed 34-antibody panel, full-Fv repair reduced final classical-TAP risk flags by up to 82\%, whereas framework only repair reduced it by upto 31\%. Developability gradients were used to steer a diffusion-based inverse-folding model, enabling it to clear 32\% of native classical-TAP risk flags. Finally, in a de novo design pipeline, trajectories generated with developability guidance carried 58\% fewer risk flags on average at the backbone-hallucination stage. Overall, we show that developability-driven guidance can be used to directly optimise antibody sequences at various stages of the design pipeline.