Therapeutic monoclonal antibodies bind to specific regions of proteins called epitopes, which elicits cellular responses. Traditional antibody discovery processes require laborious and expensive screening experiments, so computational approaches that predict epitopes and accelerate antibody discovery are in high demand. Structure-based antibody design is also important to the modern drug discovery and development process. This approach requires a high-resolution quaternary (3D) protein complex structure, whose experimental determination is often a slow process that is not always successful. Protein structure and binding interface prediction algorithms are poised to impact human health by accelerating the construction of high-confidence structural models of drug targets and biopharmaceuticals, which will help identify new therapeutic strategies. However, the current algorithms are very limited in their ability to predict high-resolution antibody-antige models, which is preventing the discovery of broad classes of therapeutics. In addition, technologies are needed to predict if a candidate antibody will fail as early as possible in the development process. With improvements in accuracy and usability, computational antibody structure and epitope prediction methods can be used to lower drug development costs and focus experiments on the most promising drug candidates. DNASTAR recently released NovaFold, a commercial version of the world leading I-TASSER protein folding algorithm (Yang Zhang, U. Michigan) running on a cloud computing platform. NovaDock, our prospective protein interaction modeling product based on the up-and-coming SwarmDock algorithm (Paul Bates, Cancer Research UK), will use the same cloud infrastructure. NovaFold is proving useful to the molecular biology community; however, it is not adapted to model protein complexes like antibodies. Also, NovaFold and NovaDock currently do not model the type of structural fluctuations that are critical for antibody recognition. These enhancements could dramatically improve the predictive accuracy of the programs. We propose to create an automatic software pipeline that predicts the highest frequency of high-resolution antibody-antigen structures that are suitable for antibody screening and biopharmaceutical design projects. Previously in Phase I, we successfully created the most accurate models for predicting epitopes by incorporating both protein sequence information and structural features derived from experimental and high- resolution predicted protein antigen structures. In this Phase II project, we will combine our fiel-leading epitope prediction models with improvements to NovaFold and NovaDock that will discover better, lower energy binding arrangements between an antibody and its antigen. The improvements will more accurately model the structural plasticity of an antibody by broadening the conformational diversity of the prediction process. At the conclusion of this work, we will deliver a cloud-based software product of suitable accuracy to dramatically increase the rate of selecting antibodies that specifically recognize a desired therapeutic target.
The use of monoclonal antibodies targeted against specific antigens are a critical, rapidly growing component of enhancing human health globally. Current practices for identifying appropriate antibodies are expensive and time consuming. We propose to dramatically streamline the monoclonal antibody selection process by combining and enhancing some of the best-in-class software programs and approaches currently available to create a software pipeline to aid users' ability to determine epitopes and model their binding to specific antibodies. The resulting computer simulated models will support major advancements in drug discovery for human health, animal health and related research.
