Theoretical Modelling Methods For Metalloproteins

As an important type of biological macromolecular system,metalloproteins play a key role in many life related processes.Understanding theoretically the interactions between metal ions and proteins is of great significance in explaining the relevant mechanisms.With the characteristics of simplicity and high efficiency,classical molecular force fields are widely used in related research of biomolecular systems.However,due to lack of the key quantum effects like polarization and charge transfer,it is not accurate enough for metalloproteins.The QM calculations are usually more rigorous and offer the capability to accurately model a range of chemistries and chemical environments in principle than MM methods while its applications are significantly limited by its computational cost.Recently,researchers are committed to developing various improved theoretical methods for metalloproteins.In this work,it is devoted to develop efficient ab initio molecular dynamic?AIMD?simulation methods for metalloproteins.Firstly,a metal molecular fractionation with conjugate caps?Metal-MFCC?is developed for efficient linear-scaling quantum calculation of potential energy and atomic forces of metalloprotein.In this approach,the potential energy of a given protein is calculated by linear combination of potential energies of the neighboring residues,two-body interaction energy between non-neighboring residues that are spatially in close contact and potential energy of the metal binding group.The present Metal-MFCC approach is linear-scaling with a low prefactor and trivially parallelizable.The individual fragment typically contains about50 atoms,and it is thus possible to be calculated at higher levels of the quantum chemistry method.Next,to further improve the computing efficiency,a force balanced molecular fractionation with conjugate caps is developed.In this approach,the amine and formyl group from the peptide bond are used as molecular caps and the polarizable multipole-based AMOEBA force field are employed to describe the electrostatic polarization effect.In this way,FB-GMFCC strikes a better balance between computing efficiency and accuracy of atomic forces.An 110ps AIMD simulation was also performed for a relatively large protein with 56 residues and 862 atoms in explicit water,the results indicate that AIMD simulation generally describes the intra-protein interactions more accurately.Then,in order to break the computational cost limitation of QM calculations for metalloproteins,an ab Initio based neural network potential model?NN/MM-RESP?is proposed to perform molecular dynamics study of zinc ion in liquid water.The predicted energies and atomic forces from NN potential show excellent agreement with the quantum chemistry calculations and its efficiency is orders of magnitude faster than that of QM calculations and QM/MM calculations.To automatically build the neural network potential model,an enhanced self-organizing incremental high dimensional neural network?E-SOI-HDNN?and corresponding training algorithm are proposed.There are three important features:?1?automated construction of reference dataset with little human intervention and low redundancy,?2?self-verification,?3?automated optimization of neural network structures.The algorithms are integrated into an open source platform named“ESOI-Chem”which could achieve auto-training of neural network potentials and different level of ab initio-based MD simulations.Finally,combined with NN/MM-RESP and E-SOI-HDNN model,ab initio based neural network potential model NN/MM-RESP-Metal is developed to achieve nanosecond molecular dynamics simulations for zinc containing metalloproteins.Among four main coordination modes of zinc-containing metalloproteins,the distribution of bond lengths and angles between zinc ion and coordination atoms show excellent agreement with the statistics in PDB Bank database.The neural network approach used in this study can applied to construct potentials to study macromolecular systems containing other ions like Cu^2+.,Fe³⁺.