Integrative modelling of the antibody antigen interactions

Background Over the years the use of monoclonal antibodies (mAbs) for therapeutic purposes has been experiencing a significative boost. The reasons behind this enormous growth reside in both their inherent structural properties and in technological advances for their production and characterisation. Indeed, their intrinsic high affinity and specificity toward a specific antigen, together with their modular anatomy, which largely simplify their engineering, are the most important drivers of this great market expansion. Antibody-based therapeutics are developed via well-established processes that can be broadly categorized into Lead Identification and Lead Optimization. During Lead Identification animal immunization or surface display technologies are used to generate a large number of ‘hit’ molecules which need to be further triaged. Following various rounds of screening and design during Lead Optimization, a small number of high affinity lead candidates are selected. During Lead Identification and Optimization, molecules are assessed for unfavourable characteristics such as immunogenicity or poor biophysical properties and experimental methodologies have been developed in order to improve such proprieties. A sound knowledge of the antibody binding site and the relationship between antibody and antigen residues is of paramount importance for the effective design of such strategies. Structural experimental techniques such as Nuclear Magnetic Resonance (NMR) or X-ray crystallography can be used to study antibody-antigen interactions but they are usually expensive and time consuming and not always applicable. Therefore, the development of computational methods is offering an attractive alternative and a faster approach for the characterization of antibodies and their interactions. Among those, docking approaches offer a valuable strategy to elucidate the interaction between antibodies and antigens providing a tool to understand the role played by each residue in the binding. Despite great progresses in protein-protein docking in general, modelling of antibody-antigen complexes, which is a specialized application of the broader field of molecular docking, has been demonstrated to remain challenging. Aim This work aimed at investigating how information about the antibody paratope and the antigen epitope can be successfully used into an information-driven docking algorithm to characterize the molecular interactions between antibodies and antigens. In order to be able to use more accurate information about the antibody binding site, this work also aimed at improving proABC, a paratope prediction tool, and at demonstrating how this information impacts antibody-antigen molecular docking. The overarching aim is to deepen our understanding of antibody- antigen recognition process and to provide computational tools and strategies that could facilitate antibody design and engineering. Results In this work I describe how information on the antibody hypervariable loops and the binding epitope can be effectively used to drive the modelling of their interaction by docking. In particular, I compare the accuracy of four docking software: ClusPro, HADDOCK, LightDock and ZDOCK in predicting antibody-antigen complexes. HADDOCK, which applies a purely data-driven strategy, performs better than the other systems especially when information about the epitope is provided.