Various coarse-grained (CG) models have become increasingly common in studies of antibody-antibody interactions in solution. These models appear poised to enter development pipelines in the near future to help predict and understand how antibody-antibody interactions influence the suitability of a given monoclonal antibody (mAb) for mass production and delivery as an antibody therapy. This blog post is a non-exhaustive summary of some of the highlights I found during a recent literature search.
Computational and theoretical models of antibody solutions can be broken down into three broad categories: (i) predictive models for the properties of dilute solutions (~10-20 mg/mL), (ii) predictive models for the properties of concentrated solutions (~100-150 mg/mL), and (iii) empirical models that fit experimental data to extract information about structures in solution.
In Case (i), the focus is frequently on predicting the second virial coefficient, which is used experimentally as a metric for dimer interactions in solution. Experimental assays of the second virial coefficient are frequently used in the mAb development pipeline to assess whether or not strong antibody-antibody interactions are present, suggesting that the mAb may suffer from developability problems down the line (e.g., aggregation, sensitivity to pH or salt). However, these experiments are consumptive of mAb material at an early stage of development when little may be available. Computational methods like those of Roberts and co-workers (Ref. 1) have proven successful in predicting the second virial coefficient for various mAbs with different responses to the change in salt content of the solution. Their simulations demonstrate that CG models with as few as twelve sites representing the full-length antibody can provide strong predictions for diverse mAbs if they account for the specific locations of charged patches on the mAb surface.
Case (ii) concerns predicting the properties of solutions of mAbs at suitably high concentration for delivery as an injection. Limits on the injection volume dictate that the mAb must be stable at concentrations of up to 150 mg/mL. Trout and co-workers (Ref. 2), in a follow-up to a previous study (Ref. 3), demonstrated that by parameterizing Fv- and Fc-specific interaction terms they are able to achieve better agreement with experimental relative viscosity measurements. Trout and co-workers also point out an interesting simulation artifact: due to the manner in which viscosity is computed from the CG simulations, its value can depend on the finite size of the simulation box. This violates our basic assumptions about how intensive properties like viscosity should behave and presents an interesting challenge moving forward.
Finally, Case (iii) studies focus on using extensive experimental information from Small Angle X-ray Scattering (SAXS) (or other scattering techniques) to parameterize CG models that can then be used to solve the inverse problem of which structures give rise to the experimental signal. In the work of Dear et al. (Ref 4.), best-fit simulations to SAXS data were used to reveal the cluster size distributions of two mAbs. Analysis of these simulations indicated that even when a CG model only has three interaction sites capable of attractive interactions, simulations detect more than three neighbors present, with multiple mAbs simultaneously interacting with the same interaction site. These results suggest that previous theoretical models that assume binding is limited to one per site may need revision (e.g., Ref. 5).
If you are interested in learning more, I suggest starting with Ref. 1 as an example of a recent paper in the area that covers various different CG models and their use cases.
References
(1) Shahfar, H.; Forder, J.K.; Roberts, C.J. J. Phys. Chem. B. 2021
(2) Lai, P.K.; Swan, J.W.; Trout, B.L. mAbs. 2021
(3) Wang, G.; Varga, Z.; Hofmann, J.; Zarraga, I.E.’ Swan, J.W. J. Phys. Chem. B. 2018
(4) Dear, B.J; … ; Truskett, T.M. J. Phys. Chem. B. 2019
(5) Schmit, J.D.; He, F.; Mishra, S.; Ketchem, R.R.; Woods, C.E.; Kerwin, B.A. J. Phys. Chem. B. 2014
