Abstract
Therapeutic antibodies are widely used for disease detection and specific treatments. However, as an exogenous protein, these antibodies can be detected by the human immune system and elicit a response that can lead to serious illnesses. Therapeutic antibodies can be engineered through antibody humanization, which aims to maintain the specificity and biological function of the original antibodies, and reduce immunogenicity. However, the antibody drug effect is synchronously reduced as more exogenous parts are replaced by human antibodies. Hence, a major challenge in this area is to precisely detect the epitope regions in immunogenic antibodies and guide point mutations of exogenous antibodies to balance both humanization level and drug effect. In this article, the latest dataset of immunoglobulin complexes was collected from protein data bank (PDB) to discover the spatial features of immunogenic antibody. Furthermore, a series of structure descriptors were generated to characterize and distinguish epitope residues from non-immunogenic regions. Finally, a computational model was established based on structure descriptors, and results indicated that this model has the potential to precisely predict the epitope regions of therapeutic antibodies. With rapid accumulation of immunoglobulin complexes, this methodology could be used to improve and guide future antibody humanization and potential clinical applications.