There is a long standing need for computational tools to predicting protein-protein interactions that can be used to mitigate physical stability of therapeutic proteins.  The present work focused on a recently developed coarse-grained molecular simulation approach to predict the experimental second osmotic virial coefficient (B<sub>22</sub>) for monoclonal antibody solutions as a function of pH, ionic strength, and the protein sequence. Static and dynamic light scattering (SLS and DLS) experiments illustrated very strong electrostatic attractions for a given wild-type (WT) monoclonal antibody in low and intermediate ionic strength conditions. The course-grained model (1bC/D, for one-bead-per-charged-site-and-per-domain) along with a biased sampling Monte-Carlo (MC) simulation was used to identify the most influential charge residues governing electrostatic interactions. This was used to identify all possible single charge-swap and charge-to-neutral mutations potentially impacting self-association of the WT MAb.  Twelve variants were made and tested experimentally at six positions by charge-swap and charge-to-neutral substitutions based on the ranking of the B22 scores and their potential impact on binding affinity. SLS and DLS experiments were used to characterize the behaviors of variants under low and intermediate ionic strength conditions. Differential scanning calorimetry (DSC) as well as circular dichroism (CD) were used to assure minimal impacts on conformational stability and secondary structure of variants. Overall, the 1bC/D model simulations provided quantitative or semi-quantitative agreement with the experimental results, and identified multiple variants that were dramatically more physically stable than WT.

Coarse-grained models for rational design of monoclonal antibodies via control of electrostatic protein-proteiniInteractions

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.