Computational modeling of antibody-antigen interactions
Benchmark sets containing the crystal structures of protein complexes and the unbound structures of the binding partners have been critical in the development of computational methods to model protein:protein interactions. Using these benchmarks, several protein:protein docking algorithms have been developed and tested to predict the 3D structure of protein complexes using the conformations of the individual partners. Despite great strides in this field, there remains room for improvement in the modeling of antibody:antigen complexes. In this work, we present an updated structural benchmark for antibody:antigen docking. The benchmark consists of 291 unique complexes curated from the Protein Data Bank (PDB) and features > 150 unique antigens, representing a substantial increase over similar existing resources. The complexes in the benchmark represent a diverse range of interfaces with respect to size, electrostatic character, and conformational change upon binding. Using this updated benchmark, we propose several improvements to improve antibody:antigen docking performance relative to existing methods.