Project B3: Coarse-graining of solvent effects in force-probe molecular dynamics simulations

The study of the conformational kinetics of biomolecules and supramolecular complexes using molecular simulations often is complicated by the fact that these processes are very slow. Various simulation techniques have been developed in order to resolve this issue. One very efficient way to investigate the atomistic details of conformational changes is provided by force-probe molecular dynamics (FPMD) simulations. In the most common realization of this technique, one end of the (supra)molecular system under consideration is fixed in space and the other end is pulled apart with a constant velocity via the application of a harmonic potential. From the distributions of the forces needed to unfold the system important information regarding the kinetics and the thermodynamics of the relevant conformational rearrangements can be obtained via a statistical analysis. The direct comparison to the results of experimental realizations of force spectroscopy usually is not easy due to the very different pulling velocities. Typically, these are about 5 orders of magnitude larger in FPMD simulations than in experiments. Therefore, every speed-up of FPMD simulations is highly desirable. A common procedure to speed up such calculations considerably is to use coarse-grained (CG) models. However, the details of the unfolding pathways cannot be monitored due to the loss of atomistic resolution.

In the present project we use hybrid schemes that allow to treat the solute in an all-atom (AA) manner and employ CG procedures for the solvent molecules. Here, the idea is to considerably reduce the computational effort needed to treat the solvent. This is of particular importance in FPMD simulations because one needs to use large simulation boxes such that also the elongated solute can be described properly. In the first funding period we used a scheme in which the coupling between the AA-solute and the CG-solvent molecules is achieved via virtual interaction sites located at the center of mass of a certain group of atoms of the solute. In the simulations, we used a system that had been extensively studied in AA simulations before, a dimer of calix[4]arene catenanes dissolved in mesitylene. The two monomers are linked by four intercalated aliphatic loops consisting of 14 methylene groups each and preventing the complete dissociation of the dimer. The
system can be viewed as a two-state model with a closed state and an open state, both of which are stabilized by a ring of 16 resp. 8 hydrogen bonds. The CG forces at the virtual sites located on the calixarene monomers were computed and then distributed uniformly among the contributing atoms. This way, a reparameterization of the force-fields is not required. The methodology had previously been shown to yield excellent results in equilibrium simulations. However, in the particular non-equilibrium situation encountered in FPMD simulations, we found qualitative but not quantitative agreement with AA-simulations of the identical system. For different choices of the virtual sites regarding both, the number and the location in the calixarene monomer, the results were very similar. An important conclusion of the study was that the impact of the solvent molecules surrounding the pulled solute cannot be treated quantitatively with the hybrid scheme used.

In the second funding period we therefore employed another hybrid scheme, the so-called adaptive resolution scheme (AdResS) which is also the subject of project B4 in the TRR 146 and which has been applied very successfully to very different situations, including non-equilibrium simulations. The calixarene dimer was placed in the centre of the AA region, the radius of which was varied. Outside this region there is a hybrid region of fixed slab thickness and the outer part of the simulation box is filled with CG solvent. As in the hybrid scheme used earlier, the CG potentials were computed using the iterative Boltzmann inversion method. In our simulations we investigated the dependence of the results on the various parameters defining the pulling procedure. Apart from the force constant of the pulling potential the most important one is the pulling velocity because this defines the hysteresis, an indicator of the non-equilibrium situation, between the pull mode and the relax mode simulations. For larger pulling velocity the hysteresis increases. Thus, it is important to investigate the performance of the AdResS for large pulling velocities. We found excellent agreement between the results of the AdResS simulations and AA simulations for all sets of pulling parameters investigated provided the AA region in the former were chosen large enough.

Furthermore, we finished our joint study with project A7 using Markov state models to implement a dynamical coarse graining procedure applied to FPMD simulations. Also, two further collaborative projects with project B5 have been finished at the beginning of the previous funding period.

In the new funding period, we plan to apply FPMD simulations using the AdResS methodology to systems that exhibit more complex folding/unfolding kinetics than the simple two-state kinetics studied up to now. In addition, we will consider systems dissolved in protic solvents because in such solvents the long-range electrostatic interactions play a more prominent role than in mesitylene. The systems to be studied will include polypeptides and also RNA and/or DNA systems.

We plan to intensify our various collaborations with other projects of the TRR. Regarding important technical details of the AdResS methodology, such as the implementation of non-spherical AA regions for the study of the unfolding of chain-like molecules our cooperation with B4 will be very important. In addition, we plan to continue our ongoing collaboration with A7 and we plan to apply our dynamic CG technique to the unfolding of a model RNA aptamer, for which we have conducted AA FPMD simulations already. Recently, we started another joint study with B5 in which the strength of hydrogen bonds stabilizing helical structures of peptides are investigated employing QM/MM methods. We plan to extensively use QM/MM in FPMD simulations in order to answer the question regarding the quality of classical force fields in non-equilibrium and to study the details of hydrogen bonding in such situations.

Force probe simulations using an adaptive resolution scheme
Marco Oestereich, J Gauss, Gregor Diezemann
Journal of Physics: Condensed Matter, (2021)
see publication


Supramolecular Packing Drives Morphological Transitions of Charged Surfactant Micelles
Ken Schäfer, Hima Bindu Kolli, Mikkel Killingmoe Christensen, Sigbjørn Løland Bore, Gregor Diezemann, Jürgen Gauss, Giuseppe Milano, Reidar Lund, Michele Cascella
Angewandte Chemie International Edition 59 (42), 18591-18598 (2020)
see publication


Force-dependent folding pathways in mechanically interlocked calixarene dimers via atomistic force quench simulations
Ken Schäfer, Gregor Diezemann
Molecular Physics 118 (19-20), e1743886 (2020)
see publication


Mechanical and Structural Tuning of Reversible Hydrogen Bonding in Interlocked Calixarene Nanocapsules
Stefan Jaschonek, Ken Schäfer, Gregor Diezemann
The Journal of Physical Chemistry B123 (22), 4688-4694 (2019)
see publication


Temperature dependent mechanical unfolding of calixarene nanocapsules studied by molecular dynamics simulations
Stefan Jaschonek, Ken Schäfer, Gregor Diezemann
The Journal of Physical Chemistry B123 (22), 4688-4694 (2019)
see publication


Temperature dependent mechanical unfolding of calixarene nanocapsules studied by molecular dynamics simulations
Takashi Kato, Ken Schäfer, Stefan Jaschonek, Jürgen Gauss, Gregor Diezemann
The Journal of Chemical Physics 151 (4), 045102 (2019)
see publication


Hybrid Particle-Field Molecular Dynamics Simulations of Charged Amphiphiles in an Aqueous Environment
Hima Bindu Kolli, Antonio de Nicola, Sigbjørn Løland Bore, Ken Schäfer, Gregor Diezemann, Jürgen Gauss, Toshihiro Kawakatsu, Zhong-Yuan Lu, You-Liang Zhu, Giuseppe Milano, Michele Cascella
Journal of Chemical Theory and Computation 14 (9), 4928-4937 (2018)
see publication


Dynamic coarse-graining fills the gap between atomistic simulations and experimental investigations of mechanical unfolding
Fabian Knoch, Ken Schäfer, Gregor Diezemann, Thomas Speck
The Journal of Chemical Physics 148 (4), 044109 (2018)
see publication


Structural Origin of Metal Specificity in Isatin Hydrolase from Labrenzia aggregata Investigated by Computer Simulations
Lalita Uribe, Gregor Diezemann, Jürgen Gauss, Jens Preben Morth, Michele Cascella
Chemistry - A European Journal l24 (20), 5074-5077 (2018)
see publication


Intramolecular structural parameters are key modulators of the gel-liquid transition in coarse grained simulations of DPPC and DOPC lipid bilayers
Stefan Jaschonek, Michele Cascella, Jürgen Gauss, Gregor Diezemann, Giuseppe Milano
Biochemical and Biophysical Research Communications 498 (2), 327-333 (2018)
see publication


Force probe simulations using a hybrid scheme with virtual sites
Ken Schäfer, Marco Oestereich, Jürgen Gauss, Gregor Diezemann
The Journal of Chemical Physics 147 (13), 134909 (2017)
see publication


Force probe simulations of a reversibly rebinding system: Impact of pulling device stiffness
Stefan Jaschonek, Gregor Diezemann
The Journal of Chemical Physics 146 (12), 124901 (2017)
see publication


Determining Factors for the Unfolding Pathway of Peptides, Peptoids, and Peptidic Foldamers
Lalita Uribe, Jürgen Gauss, Gregor Diezemann
The Journal of Physical Chemistry B120 (40),10433-10441 (2016);
see publication


Mechanical unfolding pathway of a model β-peptide foldamer
Lalita Uribe, Stefan Jaschonek, Jürgen Gauss, Gregor Diezemann
The Journal of Chemical Physics 142 (20), 204901 (2015)
see publication


Comparative Study of the Mechanical Unfolding Pathways of α- and β-Peptides
Lalita Uribe, Jürgen Gauss, Gregor Diezemann
The Journal of Physical Chemistry B119 (26), 8313-8320 (2015)
see publication