Studies To Improve Molecular Dynamics Simulation Accuracy In Protein Systems With Elastic Network Model And Polarizable Force Fields

Proteins play essential roles in a variety of life activities such as enzyme catalysis,signal transduction and substance transport,during which they undergo hierarchical conformational changes and form diverse intermolecular interactions.Molecular dynamics(MD)simulation has emerged as an effective tool for studying protein structure-dynamics-function relationships at the atomic level.It provides mechanistic basis and theoretical insight to aid experimental design and analysis.In recent years,MD simulation has evolved rapidly with the advance of high-performance computers and theoretical algorithms.However,its applications in complex biological systems are still limited by two key factors,i.e.,the sampling efficiency for essential protein conformational changes and force field’s accuracy to describe protein structural properties.Herein,I have carried out a number of projects as follows on exploring and improving the sampling efficiency and force field accuracy of MD simulations in protein systems.(1)Based on the iterative elastic network model(IterANM),I developed the IterANMIaMD enhanced sampling method to improve the sampling efficiency of MD simulations for protein conformational changes.Firstly,the IterANM method showed high sampling efficiency in protein conformation space compared with microsecond time-scale MD simulations of nine proteins systems,from which I developed the IterANM-IaMD method.I further used the Nterminal domain of calmodulin(nCaM)to evaluate the sampling efficiency of IterANM-IaMD.Results showed that the conformational space sampled by IterANM-IaMD is broader than that by Integrated accelerated Molecular Dynamics(IaMD),and the distribution of low-free-energy conformations obtained from IterANM-IaMD is also more consistent with that of experimental studies.Compared with a representative enhanced sampling method,i.e.Replica-Exchange MD(REMD),IterANM-IaMD can obtain a similar free energy distribution with only 1/4 of the computing resources,as a result of higher sampling efficiency.(2)With the IterANM-IaMD method,I explored the structural basis of protein-ligand recognition and protein conformational changes during the recognition process.From the simulations of adenylate kinase(AdK),calmodulin(CaM)and mitogen-activated protein kinase(p38α),it has shown that the IterANM-IaMD method can adequately sample the global conformational space of proteins and capture significant conformational changes during the ligand binding process,such as the cooperative movements of domains and the flipping of side chains.The intermediate-state conformations of AdK along the conformational transition path obtained by IterANM-IaMD simulations agree well with experimental structures,and the predicted binding affinity between CaM and calcium ion is more accurate than other simulation methods.Also,the obtained transition-state conformation of p38α can explain the induced-fit mechanism regulated by small molecule inhibitors.These results indicate that IterANM-IaMD can serve as an effective tool for studies of protein dynamics-function relationships.(3)I developed the IterANM-Dock docking method,in which the IterANM conformational sampling method was used in ensemble docking instead of conventional MD simulation,to improve the predictive ability for the binding mode of small molecule ligands.In the docking and post evaluation of various small molecule inhibitors of cyclin-dependent kinase 2(CDK2),it has shown that the prediction success rate of IterANM-Dock(78%)is higher than that of flexible-receptor docking AutoDockFR(67%)and ensemble docking based on MD simulation(55%)or tCONCOORD(66%)sampling method,and the computational efficiency of IterANM-Dock is higher than that of a REMD/virtual ligand filling-based ensemble docking method.In addition,IterANM-Dock also exhibits higher success rate(70%)in another docking test of 10 other different receptors.These results demonstrated the high accuracy and broad applicability of IterANM-Dock,which can play an important role in computer-aided drug design.(4)I used the AMOEBA polarizable atomic multipole force field in enhanced sampling simulations of single-channel gA and double-channel gA,and investigated its reliability in characterizing ion-transport properties of the gA(Gramicidin A)ion channel under dimerization.The influence of gA dimerization on the permeation of potassium and sodium ions through the channel was described in terms of conductance,diffusion coefficient and free energy profile.Results show that the conductance of potassium ions and sodium ions passing through the single and double channels from the polarizable force field simulation agrees well with experimental values.Further data analysis reveals the molecular mechanism of protein dimerization affecting the ion-transport properties of gA channels,i.e.,protein dimerization accelerates the permeation of potassium and sodium ions crossing the double channel by adjusting the environment around gA protein(the distribution of phospholipid head groups,ions outside the channel and bulk waters),rather than directly adjusting the conformation of gA protein.(5)In lack of general accuracy studies of polarizable force field,I studied the reliability of classical force fields(AMBER and CHARMM)and polarizable force fields(AMOEBA and Drude)to address three typical protein MD simulation challenges:protein structure refinements,protein folding and intrinsically disordered protein simulations.The simulation time for each force field is more than 40 μs.The microsecond time-scale MD simulation results show that polarizable force fields perform better in protein structure refinements and intrinsically disordered protein simulations,but their accuracy for protein folding simulations is inferior to that of classical force fields.This result provides reference information for force field application,and provides data support for the future development of high-precision force fields.