(471c) Predict the Viscosity of Concentrated Antibody Solutions Using Integrative Experimental and Computational Screening.

The stability of antibody solutions is essential for bioprocessing, manufacturing, and protein therapeutics development. The antibody stability issues are more prominent when producing concentrated solutions, which are essential for developing subcutaneous injections. For example, some antibodies exhibit high viscosity at high concentrations, presenting challenges for manufacturing, storage, and delivery. The elevated viscosity at high concentrations is due to the formation of higher-order clusters, which are driven by protein-protein interactions (PPI). However, the PPIs at concentrated concentrations are difficult to measure. Therefore, developing computational or high-throughput experimental methods that can screen antibody viscosity in the early stage is desired. In this talk, I will discuss integrative computational and experimental approaches to investigate the mechanisms and predict antibody viscosity. For the computational method, our group has implemented machine learning and molecular simulation approaches to predict antibody viscosity and PPIs. Our group collaborated with AstraZeneca to measure 229 high-concentration viscosity data, the largest to date, and developed a deep-learning model to predict viscosity. In addition, we applied a multiscale docking-based Monte Carlo simulation to predict antibody self-interactions, which is essential for their viscosity behavior. For the experimental methods, our group implemented small-angle X-ray scattering to investigate the solution characteristics of antibody solutions and developed a high-throughput screening method. The techniques developed can also be applied to study antibody-excipient interactions for formulation design.