Antibody-drug conjugates (ADCs) represent a powerful therapeutic approach for the treatment of a range of cancers. They merge the toxicity of known chemical agents with the specificity of monoclonal antibodies, thereby maximizing efficacy while minimizing adverse side effects. Although multiple ADCs have made it to the marketplace, their development remains a challenge in part owing to the lack of three-dimensional (3D) structural information that must account for the inherent flexibility of monoclonal antibodies as well as that of the drug payloads. This Perspective discusses computational methods, including machine learning and physics-based approaches, that could facilitate the interpretation of experimental data, make predictions on optimal solutions concerning drug conjugate linker type, conjugation sites and drug/antibody ratios and minimize the number of design iterations during ADC development. We explore examples of how the information content from physics-based 3D molecular modeling and simulations on model ADCs may facilitate ADC design.