Targeted protein degradation (TPD) takes advantage of the cell’s own quality-control machinery to eliminate disease-relevant proteins. PROTACs—bifunctional molecules that recruit an E3 ubiquitin ligase to the protein of interest—trigger ubiquitination and subsequent proteasomal degradation. Antibody-based degraders offer a genetically tunable alternative, with the potential for higher target affinity, modularity, and improved degradation efficiency. Designing such bispecific antibodies for membrane proteins remains difficult. Membrane proteins are conformationally dynamic, often unstable outside the lipid bilayer, and challenging to screen experimentally for high-affinity binders that can effectively recruit an E3 ligase. Deep-learning–based structure prediction and protein design now provide a way to explore bispecific formats in silico before committing to costly experiments. These methods enable us to evaluate whether a given format can simultaneously engage the target and the E3 ligase, and to identify mutations that stabilize the ternary interface. By integrating deep-learning design with membrane-embedded simulations, the project aims to narrow the experimental search space and accelerate the discovery of high-affinity antibody-based degraders.