X-ray Diffraction Data-assisted Structure Searches And Protein Structure Prediction Based On CALYPSO Method

As an efficient and reliable structure prediction method and its same-name software package,CALYPSO has been used by 1800 registered users,distributed in56 different countries or regions.Nowadays,CALYPSO method can predict the structures of 3D crystals,2D layers,2D surface reconstructions and 0D clusters.And it can also design novel functional materials,such as superhard materials.Based on the CALYPSO method,previously developed by our team,we develop the X-ray diffraction data-assisted structure search and protein structure prediction methods.As an important experimental technique,X-ray diffraction?XRD?can determine the crystal structure at the atomic level.How to solve the crystal structure precisely from the XRD data is one of key issues in practice.Based on the CALYPSO method,we have developed the X-ray diffraction data-assisted structure search method to solve the crystal structure.It's difficult to solve crystal structure from powder diffraction patterns,because the three-dimensional information is compressed into one dimension leading to peak overlap in the powder diffraction pattern.Such peak overlap limits obtaining the peak positons and intensities.To overcome the peak overlap problem,the direct-space method is developed for structure solution without extracting the peak intensities.The essence of the direct-space method is exploring the hypersurface to locate the global minimum.In the direct-space method,larger numbers of trail structures are generated and simulated the diffraction pattern for each trail structure.The suitability of each trail structure is assessed by the similarity between simulated XRD pattern and experimental pattern.The similarity is frequently quantified by a figure-of-merit(R_wp).And the most similar one is selected as the final structure.With the idea of the direct-space method,we have developed an X-ray diffraction data-assisted structure search method to solve the crystal structure on the basic of CALYPSO method.Several techniques are implemented to improve the efficiency of structure search,including background subtraction,simulated annealing algorithm to optimize profile parameters,quantifying the similarity between two patterns by a weighted cross-correlation function and the CALYPSO structure search method.Our method is benchmarked by two known systems?hexagonal ZnO and Anatase TiO₂?.Meanwhile,we search the structures of two unknown systems?CaLi₂ and Ca₃C₂?with the XRD data by our method.Xie et al.preformed global structure searching under guidance of the energy and reported two CaLi₂ structures with space groups C2/c and P2₁/c in the range of 35-105 GPa.But the simulated XRD patterns don't agree with the experimental one.After the XRD data-assisted structure search,we find a C222₁-CaLi₂ structure,whose simulated XRD pattern excellently agrees with the experimental one.For Ca₃C₂,the simulated XRD pattern of previously reported P4/mbm-Ca₃C₂ doesn't agree with the experimental one.We propose a Cm2m-Ca₃C₂structure by our method,whose simulated XRD pattern agrees with the experimental one.We believe that the Cm2m-Ca₃C₂ will be a good candidate for the synthesized binary compound Ca-C system.Protein is the building block of life and governs abundant life processes.X-ray crystallography and nuclear magnetic resonance are main technologies to determine the protein structures.These two methods are labor-intensive and expensive.So the development of an efficient computer-aided method to predict protein structure is very desirable.According to Anfinsen thermodynamic hypothesis,the native conformation lies at the global free energy minimum.The basic of protein structure prediction is exploring the free energy surface.We develop a database assisted ab initio protein structure prediction method via swarm intelligence algorithm.The method has been implemented in the CALYPSO software package.ab initio method doesn't depend on any known protein structure database and searches the free energy surface under the guidance of energy function.PSO algorithm,we have developed an ab initio protein structure method.In our method,the ideal bond lengths and angles are used to construct protein conformations and the dihedral angles are only variables for different conformations.The secondary structure constraint is implemented,which can substantially reduce the search space and improve the efficiency of our method.Our method is benchmarked by distinct folding peptides with different sizes?small proteins with the length less than 50 amino acids?,including loop peptides,?peptides and?peptides.Native-like structures are successfully predicted,which demonstrates the reliability of our method.With the increasing of protein size,the conformational phase space becomes extremely complicated and the number of local minia sharply increases.So the ab initio method is not suitable for large proteins.But large number of proteins in the database can be used as templates to construct the structures for large proteins.So we develop a database assisted protein structure prediction based on the ab initio method,with the introduction of protein structure database.Our method can construct the initial conformation by the template proteins in the database and build the fragment library from the templates to update protein conformations.Meanwhile,our method adopts three methods for structure evolution:improved PSO algorithm,random perturbation and fragment substitution.Our method is benchmarked by 32 distinct folding proteins,including?proteins,?proteins and??proteins,with the length from47 to 118 residues.The native-like protein models are all produced,which are useful for biological application.Especially,seven high-resolutions models are generated,with excluding homologous proteins.It demonstrates the accuracy and robustness of our method.And we compare the predicted models by our method with ones predicted by ROSETTA method and I-TASSER method,which are widely used for protein structure prediction.The precision of most models is comparable to or even better than ones predicted by ROSETTA method or I-TASSER method.It indicates that our method has the powerful ability and can be used to predict protein structures.