QM/MM metadynamics of thiol-disulfide exchange using a neural network correction

  • Date:

    July 12

  • Speaker:

    Denis-Mario Maag (RTG 2450, KIT, Germany)

  • Time:

    10:10 - 10:30

  • Abstract

     

    Denis Maag, Claudia L. Gomez Flores, Mayukh Kansari, Tomáš Kubař and Marcus Elstner

     

    Karlsruhe Institute of Technology (KIT)

     

    Molecular dynamics and free energy calculations of chemical reactions in which bonds are broken and formed require a quantum mechanical (QM) description. However, the application of most QM methods is restricted to very short time scales (femtoseconds to picoseconds), due to their high computational cost.

    With the semi-empirical Density-Functional Tight Binding (DFTB) method, which is up to 1000x faster than DFT, it is possible to simulate such reactions on longer time scales. However, some classes of reactions suffer from qualitative and quantitative errors due to approximations in DFTB.

    In thiol-disulfide reactions [1], a nucleophilic substitution that occurs in a large class of proteins, the transition state is inaccurately described with DFTB [2]:

    (i) the sulfur-sulfur distances in the transition state are too long

    (ii) the transition state geometries exhibit shallow minima on the free energy landscape, rather than saddle points.

    Both artifacts have been corrected by applying a machine learned energy correction with respect to CCSD(T) level of theory [3].

    The additional computational effort is comparable to that of DFTB, allowing extensive sampling of thiol-disulfide exchange with ab initio accuracy at low computational cost.

    The free energy profiles of thiol-disulfide exchange in a small model system and a blood protein were obtained with QM/MM metadynamics using the machine learned energy correction.

    In a next step, the computational efficiency may be further increased by using a neural network instead of DFTB, which learns the energies directly as well as the charges of the QM atoms.

     

    [1] Neves, R. P. P.; Fernandes, P. A.; Varandas, A. J. C.; Ramos, M. J. Benchmarking of Density Functionals for the Accurate Description of Thiol–Disulfide Exchange. J. Chem. Theory Comput. 2014, 10, 4842– 4856

    [2] Putzu, M.; Gräter, F.; Elstner, M.; Kubař, T. On the Mechanism of Spontaneous Thiol–Disulfide Exchange in Proteins. Phys. Chem. Chem. Phys. 2018, 20, 16222– 16230

    [3] Gómez-Flores, C. L.; Maag, D.; Kansari, M.; Vuong, V. Q.; Irle, S.; Gräter, F.; Kubař, T.; Elstner, M.; Accurate Free Energies for Complex Condensed-Phase Reactions Using an Artificial Neural Network Corrected DFTB/MM Methodology., J. Chem. Theory Comput. 2022. 18, 2, 1213–1226