Protein-protein interactions

State of the art and preliminary work.

Research at the Institute of Molecular Modeling and Simulation has a strong focus on the development of more efficient methods for tackling some of the remaining challenges in the prediction of the relative binding free energies of protein-ligand complexes. However, these methods are only of limited use to quantify the interactions in protein-protein complexes. It is clear that alchemical approaches to compute the full binding free energy of two proteins is unlikely to work, as this would require the complete annihilation of an entire protein. The required sampling to accurately perform such a process is simply inaccessible, even for the largest supercomputers currently available.
Rather, one can try to simulate the actual binding process. Ideally, a straightforward MD simulation readily samples the reversible binding equilibrium and the binding free energy can be obtained by direct counting of the bound and unbound states. While this again seems feasible for some small molecules it is unlikely to work for protein-protein interactions. More commonly, one forces the binding process along a judiciously chosen reaction coordinate.
Ideally, one would simulate the binding process reversibly, i.e. simulate both the unbinding process as well as the binding process. Dissipation effects then result as a strong hysteresis between the two processes. However, bringing protein structures together is not straightforward and the correct binding pose needs to be found. Merely enhancing the sampling and selection of an appropriate reaction coordinate will not be sufficient to tackle the sampling issues in protein-protein interactions.

Objectives.

In this PhD project, we will address both the sampling and the scoring issue for protein-protein interactions. We have recently developed GroScore as an efficient means to identify the correct binding orientation out of a previously generated set of suggested poses. We will further refine this approach and combine path-sampling methods with alchemical modifications to also allow for a quantitative description of such interactions. We will selectively remove the interactions between the two proteins only. The intramolecular interactions as well as all protein-solvent interactions will remain unaltered. This will involve a significantly smaller amount of interactions, and free-energy estimates can be converged. To avoid the two binding partners occupying the same space, we will further reduce this to only remove the attractive components of the interactions, while maintaining the repulsive ones. This will facilitate the unbinding of the two proteins along a predefined path.

Anticipated outcomes and technological innovation.

An efficient quantification of protein-protein interactions is one of the major challenges of free-energy calculations in biomolecular simulation. Many of the pharmaceutics of the future will be based on biologics, such as antibodies or other neutralizing entities. Furthermore, to optimize PROTAC designs, an optimization of protein-protein interfaces will become possible. Reliable prediction of protein-protein complexes and the quantification of their interaction strength, will furthermore increase our understanding of the stability and aggregation propensities of biotechnological products. Taken together, this project opens the way for a wide range of future developments.

J.W. Perthold and C. Oostenbrink. Simulation of reversible protein-protein binding and calculation of binding free energies using perturbed distance restraints. J. Chem. Theory Comput. 13, 5697 - 5708 (2017)
J.W. Perthold and C. Oostenbrink. GroScore: Accurate scoring of protein-protein binding poses using explicit-solvent free-energy calculations. J. Chem. Inf. Model. 59, 5074 - 5085 (2019)