B03 - Multilevel coarse graining of multiscale problems
Head(s): Prof. Dr. Cecilia Clementi (FU Berlin), Prof. Dr. Beate Koksch (FU Berlin), Prof. Dr. Roland Netz (FU Berlin), Prof. Dr. Christof Schütte (FU Berlin, ZIB)
Project member(s): Cihan Ayaz, Dr. Andreas Bittracher, Laura Lavacchi, Andrea Gulyas, Alexander Sikorski, Benjamin Hery, Artem Pavlov
Participating institution(s): FU Berlin
We will develop coarse-graining methods at different resolution, that preserve the kinetics of rare events in molecular systems and apply them to characterize the aggregation and filament growth in amylogenic peptides. Towards this goal, we will continue to develop the transition manifold (TM) framework, a dimensionality reduction method that finds optimal reaction coordinates by learning the geometrical structure of transition probability densities in complex systems. We plan to integrate this method with the multi-dimensional generalized Langevin equation (GLE) with non-linear friction into a truly multilevel approach and to merge it with molecular coarse graining procedures based on force matching and machine learning. As application we will extend our designated benchmark problem, which is peptide aggregation, and study the process of simultaneous co-folding and docking of a protein to a fibril structure. As a specific example the islet amyloid polypeptide (IAPP), a 37-residue peptide hormone, will be used, which in preliminary atomistic MD simulations is shown to form stable fibrils. The relevant reaction coordinates will be identified using TM methods and then be employed within a multidimensional GLE to extract free energies and memory function matrices from simulation trajectories. For simulating fiber-growth and fiber-breakage events on timescales longer than amenable to atomistic simulations, molecular (mesoscale) coarse-grained models will be developed and the corresponding free energies and memory function matrices will be compared to the ones obtained from atomistic models. As a main result, we will extract and analyze the off-diagonal memory functions that reveal the kinetic cross-coupling between reaction coordinates and thereby be able to describe the kinetic multi-scale coupling between inter- and intra-peptide folding. In close dialogue with theory, experiments will shed light on the protein fibril elastic and dynamic properties as well as on the crossover between ordered domains, in the interior of the fibrils, and disordered domains, which form a diffuse peptide corona that might play a role in the escape of fibrils from the protein degradation machinery. A schematic overview of the project is shown in Figure 1.