Therapeutic proteins, including antibodies and other scaffolds such as DARPins and affibodies have become important drugs to treat cancer, infectious and cardiovascular diseases, arthritis, inflammation and immune disorders, and are expected to drive further growth of the biotechnology industry. Although the main tool for discovering high-affinity binders for target proteins is the use of combinatorial libraries in various display platforms, recent studies demonstrate the potential of computational approaches to obtaining models of antibody-antigen complexes and to the design of therapeutic proteins. However, these studies also show that there is space for substantial improvements. The goal of this proposal is developing improved software (1) for determining the structure of antibody-antigen complexes by integrated homology modeling and docking, and (2) for interface optimization to improve binding affinity, primarily targeting the design of therapeutic proteins with non-antibody scaffolds, because the latter have the majority of variable amino acids in the interface. The methodology is based on two key ideas. First, the modeling of the antibody H3 loop will be fully integrated with docking, i.e., structures will be docked first without the loop, and in each docked structure the loop will be rebuilt by accounting for the position of the antigen. Note that this approach inverts the usual sequence of first building and then docking homology models. Another innovative element of the method is searching for optimal conformers of the interface side chains directly in the process of docking. The constructed structures will be refined using "stability analysis", based on repeated Monte Carlo minimization runs to explore energy funnels around native structures. This approach enables global docking without any information on the epitope, and provides improved accuracy. While we focus on antibodies, the methods developed here will be generally applicable to docking homology models of any protein therapeutics. Given an initial complex of a therapeutic protein (or its homology model) with a target protein, the second key idea we use is starting the optimization of the interface with determining binding hot spots on both proteins. We then directly build the critical side chains of the therapeutic protein into the hot spots to maximize the overlap with the pre-calculated density of probes, which predicts the binding affinity at each point. Since the scoring function of the design includes both a measure of overlap with the hot spots and a term describing the energy of the complex, the method accounts both for changes in the direct interactions and other changes (e.g. new hydrogen bonds or salt bridges) that affect the stability of the protein rather than the protein-protein interface. In addition to developing and validating the two methods, Acpharis will develop commercial quality programs for the biotechnology industry. The programs will be marketed both via Schr?dinger and directly to our customers, and will also be used in our consulting and contract research practice. !
Therapeutic proteins, including antibodies and other scaffolds, have become important drugs to treat cancer, infectious and cardiovascular diseases, arthritis, inflammation and immune disorders, and are expected to drive further growth of the biotechnology industry. The goal of this proposal is developing improved computational methods and software for modeling and optimal design of protein therapeutics.