A series of coarse-grained models for molecular simulation of proteins are considered, with emphasis on the application of predicting protein–protein self-interactions for monoclonal antibodies (MAbs). As an illustrative example and for quantitative comparison, the models are used to predict osmotic virial coefficients over a broad range of attractive and repulsive self-interactions and solution conditions for a series of MAbs where the second osmotic virial coefficient has been experimentally determined in prior work. The models are compared based on how well they can predict experimental behavior, their computational burdens, and scalability. An intermediate-resolution model is also introduced that can capture specific electrostatic interactions with improved efficiency and similar or improved accuracy when compared to the previously published models. Guidance is included for the selection of coarse-grained models more generally for capturing a balance of electrostatic, steric, and short-ranged nonelectrostatic interactions for proteins from low to high concentrations.

Attractive electrostatic protein–protein interactions (PPI) necessarily involve identifying oppositely charged regions of the protein surface that interact favorably. This cannot be done reliably if one only considers a single protein in isolation unless there are obvious charge “patches” that result in extreme molecular dipoles. Prior work [J. Pharm. Sci.2019108120132] identified three monoclonal antibodies (MAbs) that displayed experimental behavior ranging from net repulsive to strongly attractive electrostatic interactions. The present work provides a systematic computational approach for identifying the origin of diverse PPI, in terms of which sets of amino acids or individual amino acids are most influential, and determining if there are different patterns of pairwise amino acid interaction “maps” that result in different behaviors. The charge was eliminated computationally, one by one, for each charged residue in the wild-type sequences, which resulted in predicted changes in the second osmotic virial coefficient. The results highlight interaction “maps” that correspond to cases with qualitatively different net electrostatic PPI for the different MAbs and solution conditions, as well as key sets of residues that contribute to strongly attractive PPI. A more computationally efficient method is also proposed to identify key amino acids based on Mayer-weighted interaction energies.

Electrostatically driven attractions between proteins can result in issues for therapeutic protein formulations such as solubility limits, aggregation, and high solution viscosity. Previous work showed that a model monoclonal antibody displayed large and potentially problematic electrostatically driven attractions at typical pH (5−8) and ionic strength conditions (∼10−100 mM). Molecular simulations of a hybrid coarse-grained model (1bC/D, one bead per charged site and per domain) were used to predict potential point mutations to identify key charge changes (charge-to-neutral or charge-swap) that could greatly reduce the net attractive protein−protein self-interactions. A series of variants were tested experimentally with static and dynamic light scattering to quantify interactions and compared to model predictions at low and intermediate ionic strength. Differential scanning calorimetry and circular dichroism confirmed minimal impact on structural or thermal stability of the variants. The model provided quantitative/semiquantitative predictions of protein self-interactions compared to experimental results as well as showed which amino acid pairings or groups had the most impact.

Static light scattering (SLS) was used to characterize five monoclonal antibodies (MAbs) as a function of total ionic strength (TIS) at pH values between 5.5 and 7.0. Second osmotic virial coefficient (B 22 ) values were determined experimentally for each MAb as a function of TIS using low protein concentration SLS data. Coarse-grained molecular simulations were performed to predict the B 22 values for each MAb at a given pH and TIS. To include the effect of charge fluctuations of titratable residues in the B 22 calculations, a statistical approach was introduced in the Monte Carlo algorithm based on the protonation probability based on a given pH value and the Henderson−Hasselbalch equation. The charged residues were allowed to fluctuate individually, based on the sampled microstates and the influence of electrostatic interactions on net protein−protein interactions during the simulations. Compared to static charge simulations, the new approach provided improved results compared to experimental B 22 values at pH conditions near the pK a of titratable residues.