This project aims to establish new approaches toward large-scale simulations of biomolecular systems. We concretely seek to develop bottom-up coarse-grained models for the immature HIV CA-SP1 lattice including protease dimers as well as for RNA systems. Proteolytic cleavage of the CA-SP1 junction is an essential process for disassembly of the immature HIV lattice and hence critical to HIV infectivity. How the protease processes the immature HIV lattice and hence triggers its disassembly has remained an open question of high therapeutical interest that we address in this project. To this purpose, we have performed structural modelling of the CA-SP1/protease complex and performed extensive atomistic simulations that we subsequently used to build large-scale coarse-grained models. Development of such models that exhibit realistic affinities between the binding partners is a major difficulty that we solved by designing a new machine-learning technique. In this way, we are able to simulate the immature HIV lattice that is composed of more than 3100 proteins exposed to protease dimers. Furthermore, we have implemented a bond-cleavage mechanism for the CA-SP1 junction in the presence of protease dimers. Performing reactive, coarse-grained simulations for this system, we aim to toward a mechanistic understanding of how proteases cooperatively disassemble the immature HIV lattice. Such insight is highly relevant for the design of new drugs inhibit HIV maturation through stabilization of the immature HIV lattice such as bevirimat.
Another objective of this project is the development of a hybrid atomistic/coarse-grained (MM/CG) model for RNA. Atomistic simulations of RNA macromolecules such as the ribosome are computationally too expensive. Long-range or cooperative effects, however, often play an important role in the conformational behavior of local domains. A pure coarse-grained representation, on the other hand, is of too low resolution to address fine-grained structural changes. Thus, we seek to resolve this dilemma by developing a hybrid model in which critical parts of the system are atomistically resolved, whereas the surrounding is described by a coarse-grained model. In this way, we seek to lay the methodological foundation for a new simulation standard, that facilitates the simulation of drug binding to RNA target-sites while capturing the macromolecular environment.
