Predicting the Dynamic Viscosity of High-Concentration Antibody Solutions with a Chemically Specific Coarse-Grained Model

Abstract

The viscosity of high-concentration protein solutions is a critical parameter in bio-pharmaceutical formulation development. Conventionally, the viscosity is measured and optimized in labor-intensive experimental workflows that require a lot of material. While predicting the viscosity with atomistic molecular dynamics (MD) simulations is feasible, they are computationally prohibitively expensive due to the large system sizes and the long simulation times involved. Coarse-grained MD (CG-MD) simulations significantly reduce computational demands, but evaluating their accuracy and predictive power requires rigorous validation. Here, we assess the capability of the Martini 3 CG force field to predict the viscosity of high-concentration antibody solutions. We show that a refined Martini 3 force field, with optimized protein-protein interactions, can predict the elevated viscosities observed in concentrated solutions of F(ab)2 fragments of the therapeutic monoclonal antibody (mAb) omalizumab. Furthermore, we show that our previously developed Martini 3-exc model for arginine excipients successfully captures the trend of lowering viscosity, as observed in our rheology experiments. These findings open the way to physics-based computational prediction of the properties of dense biopharmaceutical solutions via large-scale MD simulations.