Development Of The Ultra-Coarse-Grained Model For Proteins

Protein functions are closely related to their conformational changes.Because of the complexity of conformational changes,it is difficult to obtain the information of intermediates and transition states from experimental techniques.Although All-Atom(AA)and residue-based Coarse-Grained(CG)Molecular Dynamics(MD)simulations can predict detailed dynamical data of conformational changes,their applications are still limited by the system size and the simulation timescale.And they can only be used to explore conformational changes of small to medium-sized proteins.In order to simulate larger proteins for long timescales,we have developed a UCG model with lower resolution,compared to residue-based CG models,to simulate conformational changes of proteins at the multiscale level.The main results of this dissertation are summarized as follows:(1)Because the resolution of UCG model is less than residue-based CG models,a systematic method is required to construct the protein UCG model.By systematically comparing various methods for the UCG protein division,we found that our UCG protein division method has high efficiency and accuracy,based on the secondary structure analysis and the optimal value of convergence.At the same time,by using fluctuations from AA-MD data,we found that the UCG model constructed by this method can accurately describe the details of secondary structure in H-Ras system.(2)The UCG division methods usually depend on the structural information at the AA level,but for some large biomolecules,they can only be observed in low density data from the electron microscope.Therefore,we developed a UCG division method,based on the density map from the electron microscope.Then,we validated our model with a large number of typical biological structures,such as chromosomes,viral coats and microtubules.We also parameterized our model by using an elastic network potential which is specially proposed to describe the interaction of pseudo-particles on the density map.Based on this potential,we carried Pulling MD simulations of F-Actin and Collagen,and the results showed that our UCG model can be applied to the prediction of mechanical properties of large biomolecules.(3)Because of the low resolution of UCG models,the development of UCG forcefields generally is difficult.For this reason,we have developed a structure-based forcefield that can be used to describe the anharmonic interactions near the equilibrium configuration in a single protein.We found that this forcefield can accurately reconstruct the RMSF feature of the AA model for the H-Ras protein with appropriate forcefield parameters and UCG resolutions.At the same time,the UCG model was able to reproduce the RMSF data from the AA model in relatively short time period,suggesting that the UCG model can be applied to accelerate the conformational dynamics.(4)In order to accurately describe the multi-state conformational changes of a single protein at the UCG level,we developed a multi-state structure-based forcefield and a systematic parameterization method.The mathematical expression of the force field is concise,and there are few parameters that need to be fitted.Under the resolution of UCG,this forcefield accurately reproduces the 2D free energy map and intermediates of Adenylate Kinase and other proteins,in which the structure of intermediates is highly similar to that of crystals.Accordingly,this forcefield provides a reliable expression to describe the multi-state conformational changes of a single UCG protein.(5)Finally,we proposed a semi-analytic approach to fit the Lennard-Jones(LJ)parameters of the CG model,based on the LJ potential energy surface AA model,which does not require AA-MD simulations.The test calculations for proteins and phospholipids show that this method can not only obtain LJ parameters of CG model accurately and efficiently,but also provides a scheme to optimize the protein division.Overall,this method opens a new avenue for study of protein-protein interactions using the UCG model.