Therapeutic monoclonal antibodies bind to specific regions of proteins called epitopes, which elicit cellular responses that treat or cure disease. Discovering therapeutic antibodies traditionally requires costly and labor- intensive, laboratory-based screening experiments. Computational approaches that select antibodies with the most desirable pharmaceutical properties are thus poised to improve health by accelerating the development of new drugs. Unfortunately, current algorithms are often unable to distinguish stronger-binding antibodies from weaker ones. Improvements to structure prediction and molecular visualization will lower costs and increase the speed with which new drugs are developed by allowing researchers to focus on the most promising candidates as early in the process as possible. DNASTAR’s goals are to increase the speed of predicting the structure of antibody-antigen interactions using superior mathematical methods and to transform antibodies with micromolar binding affinity into those with improved nanomolar affinity using new computer-aided antibody design techniques. This will accelerate antibody discovery by enabling detailed and accurate immune complex structure predictions and structure-based chemical liability detection at a high-throughput scale. In Phase II, we first created an in silico human germline sequence library and used it to simulate the natural V(D)J and VJ recombination events of the immune system, generating a new library of assembled antibody sequences. To select antibody candidates that bound a chosen target, we developed a simulation algorithm in which antibody candidates were docked against a chosen target protein. The 24 candidates with the best predicted binding energy were converted to single-chain antibodies and propagated in CHO cells. Three candidates were found to bind the target using native Western blots. The binding affinity and kinetics of these three candidates were then measured by bio-layer interferometry. The tightest binding candidate was then subjected to a form of simulated affinity maturation where individual site-directed mutations were ranked by their predicted ability to enhance affinity for the antigen. Four out of five tested variants showed improved binding over its parent using bio-layer interferometry. The goal of our Phase IIB proposal is to build upon this success and further improve predictive capability by incorporating unequaled algebraic mathematics and computational acceleration techniques to support the virtual screening of tens of thousands of antibody sequences. For the first time in history, this will enable antibodies to be selected for development by first modeling them from germline sequences using a “virtual immune system.” Our ultimate intent is to deliver a complete antibody discovery pipeline that is powerful, accurate, produces fast results, and yields lab-scale quantities of DNA and protein materials for the selected antibodies.