A molecular simulation method based on finite element representation

Fitting the surface electrostatic and van der Waals force field parameters of macromolecules through finite element representation method, the problem of low computational efficiency of biomolecule dynamics simulation under mesoscopic scale in the prior art is solved, and high-precision dynamics simulation is achieved, which is suitable for Study the dynamics of complex biological systems.

CN114595598BActive Publication Date: 2024-11-26DALIAN INSTITUTE OF CHEMICAL PHYSICS CHINESE ACADEMY OF SCIENCES
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011430190.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-07
Publication Date
2024-11-26
Estimated Expiration
2040-12-07

AI Technical Summary

Technical Problem

Existing biomolecular dynamics simulations are difficult to achieve efficient simulation at mesoscopic scale, especially due to the lack of appropriate models and force fields, which leads to long calculation time and huge resource consumption, making it impossible to effectively simulate complex physiological or pathological phenomena in real cells. .

Method used

Using the finite element representation method, a high-precision finite element grid is generated by fitting the parameters of the surface electrostatic electricity and the van der Waals force field of macromolecules, which are used to simulate the dynamic process of biological molecules and replace tens of thousands of irregular macromolecules with atoms. , realize dynamic simulation.

Benefits of technology

It significantly improves the simulation time scale, realizes dynamic simulation of mesoscopic scale, is suitable for studying complex processes such as molecular recognition and assembly, improves computing efficiency, and can simulate large-scale biological systems at the millisecond time scale.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114595598B_ABST
    Figure CN114595598B_ABST
Patent Text Reader

Abstract

The present invention relates to a molecular simulation method based on finite element representation, comprising the following steps: obtaining spatial point cloud data of biological macromolecules to perform surface triangular mesh division and generate finite elements; taking the geometric center of each triangular surface of the finite element outer surface as the position of the charge to be fitted; obtaining the charge quantity of the center of the finite element surface by fitting the electrostatic potential around the finite element macromolecule, and generating the electrostatic potential around the macromolecular protein; obtaining the van der Waals parameters of the center of the finite element surface by fitting the van der Waals attraction and repulsion potential energy respectively, so as to simulate the original van der Waals potential around the protein. The present invention can select appropriate finite element meshes for different systems to perform triangular mesh representation on biological macromolecules, and then fit the electrostatic van der Waals parameters on the relevant mesh surface, so as to truly realize the coarse-grained description of the adjustable scale of biological macromolecules. Its dynamic simulation process is suitable for mesoscopic scale simulation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of bioinformatics, computing and optimization methods, and specifically relates to a finite element representation of molecular dynamics simulation method, the core of which is the fitting of finite element surface electrostatic and van der Waals force field parameters. Background Art

[0002] Cells are the most important research objects in life sciences and biomedicine. Complex physical and chemical elementary processes such as molecular recognition, enzyme catalysis, and transmembrane material transport involving a large number of molecules are constantly occurring inside them, thereby maintaining the normal function of life or causing disease. In terms of experimental research, the 2014 Nobel Prize in Chemistry was awarded for pioneering research on super-resolution imaging technology, which has opened up new horizons for life science and biomedical research. Real-time imaging has broken through the optical diffraction limit in spatial resolution, reaching the nanoscale at the highest, and the temporal resolution has also reached the millisecond scale, allowing us to achieve unprecedented precision in the complex physiological or pathological processes occurring inside cells. At the same time, with the rapid development of experimental techniques for detecting protein structures such as X-ray crystal diffraction, nuclear magnetic resonance, and cryo-electron microscopy, a large number of complex biological macromolecules and their complexes have been resolved with atomic resolution. However, these experimental techniques can only provide the static structure of a single biological molecule or complex under specific conditions under non-physiological conditions, and it is difficult to obtain its true dynamic behavior related to its function, and it is impossible to explain the microscopic mechanism and molecular mechanism of the complex physical and chemical phenomena observed by super-resolution imaging experimental technology at the mesoscopic scale.

[0003] Although existing experimental techniques cannot achieve the study of physical and chemical processes with atomic space and ultrafast time resolution, molecular dynamics simulation theory and computational methods based on physical principles make it possible and have been widely used in the study of the microscopic dynamic behavior of single biological macromolecules and their functional mechanisms. At present, biomolecular dynamics simulations are all based on single molecules or simple complexes obtained from structural biology experiments as research objects. For single physical and chemical processes, theoretical research is carried out on a nanosecond to microsecond time scale in a simple physiological salt solution or membrane phospholipid component environment. However, the physiological or pathological phenomena occurring in real cells are completed by biomolecular complexes and their polymers on a larger spatial scale in a complex physiological salt solution or multi-component cell membrane environment, over a longer time scale and with multiple physical and chemical elementary processes, such as recognition, catalysis and product dissociation of enzyme substrates, phase transition processes of large-scale aggregation caused by environmental response of water-soluble proteins, self-assembly of ultra-large functional protein complexes on cell membranes, neural signal transduction processes, etc. Such complex systems usually contain more than one million atoms, and the time scale is in milliseconds or even longer. Except for the customized chip-specific machine Anton II of DEShaw in the United States, which can reach a speed of 1 millisecond in 100 days, the existing all-atom biomolecular dynamics simulations cannot achieve effective simulation. Taking the fastest GPU hardware accelerator card V100 as an example, it can only simulate a million-atom system for 20 nanoseconds a day at most. my country's fastest supercomputer TaihuLight can only run for 100 nanoseconds a day at most, which is quite different from the requirements of real biological experimental research. The dynamic simulation and mechanism analysis of biological systems in near-physiological environments and mesoscopic experimental scales (micrometer space and millisecond time) are extremely complex and difficult for theoretical computational chemistry research.

[0004] Molecular dynamics simulation at the mesoscopic scale is a hot topic today, but the current simulation methods that can be used to explore the mesoscopic scale are very limited. On the one hand, there is a lack of suitable models, and on the other hand, there is a lack of corresponding force fields. The fitting of electrostatic and van der Waals parameters for coarse-grained models of biomacromolecules plays a very critical role in the research on biomolecular structure design and molecular dynamics. Because the macromolecular system model contains thousands or even tens of thousands of atoms at atomic resolution and has a very complex spatial structure, the conventional dynamics simulation process requires a considerable amount of computing time and very huge computing resources. Therefore, it is extremely necessary to develop dynamics simulation methods and their supporting force fields for the mesoscopic scale. Summary of the invention

[0005] The content of the present invention is to provide a coarse-graining method for finite element representation of macromolecules. A small number of outer surfaces (tens or hundreds) of finite elements are used to replace tens of thousands of atoms of irregular macromolecules to generate electrostatic potential and van der Waals potential around the macromolecules, and the electrostatic and van der Waals parameters of the surface are fitted. Then the molecular dynamics simulation of the finite element grid is performed using the fitted parameters. Traditional finite element models are all theoretical modeling for macroscopic objects, including bones, heart, vascular tissue organs, etc. of biological systems, as well as numerical solutions for some mathematical problems. There are only a few literature reports on finite element dynamics simulation of biological molecular systems, and they are all based on simple and rough elastic force models to approximate the description of molecular vibration properties. The present invention proposes the construction of a high-precision finite element model.

[0006] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a molecular simulation method based on finite element representation, comprising the following steps:

[0007] 1) Use the coordinates and van der Waals radius of the biomacromolecule to obtain the spatial point cloud data of the biomacromolecule, perform surface triangulation on the point cloud data, and generate finite elements;

[0008] 2) Take the geometric center of each triangular surface of the finite element outer surface as the position of the charge to be fitted;

[0009] 3) The charge at the center of the finite element surface is obtained by fitting the electrostatic potential around the finite element macromolecules, and the electrostatic potential around the macromolecular protein is generated by the superposition principle to simulate the original electrostatic potential around the protein;

[0010] 4) The van der Waals parameters of the finite element surface center are obtained by fitting the van der Waals attraction and repulsion potential energies respectively; based on the position of the finite element center and the van der Waals repulsion and attraction parameters, the van der Waals potential around the biomacromolecule is generated through the superposition principle to simulate the original van der Waals potential around the protein.

[0011] Step 2) is as follows:

[0012] Take each external surface triangle (P i_1 ,P i_2 ,P i_3 ) as the geometric center of the charge to be fitted. i :

[0013] O i =(P i_1 +P i_2 +P i_3 ) / 3.

[0014] Step 3) is as follows:

[0015] (1) Generate the electrostatic potential around the protein using the Tinker software;

[0016] (2) Generate a distance matrix between the selected discrete points and the grid points around the irregular surface protein:

[0017] (3) Perform QR decomposition on the distance matrix to obtain the upper triangular matrix Fitting the charge of a point charge

[0018] The method of generating a distance matrix between the selected discrete points and the grid points around the irregular surface protein comprises the following steps:

[0019] For each spatial point Pos around a protein pot_i , whose potential is determined by the discrete point charges on the protein surface at Pos pot_i The superposition of the upper potential, that is

[0020]

[0021] in, R ij is the distance between the spatial grid point i around the molecule and the center of the triangular face j, spot i is the electrostatic potential value of the ith simulated value, q j is the charge at the center j of the triangle, N cg is the total number of triangle center points.

[0022] Write the solution process of all points in the form of a matrix,

[0023]

[0024] Among them, R mat is the distance matrix;

[0025]

[0026] N cg Expressed as the total number of triangle center points.

[0027] In step 4), the van der Waals parameters of the finite element surface center are obtained by fitting the van der Waals attraction and repulsion potential energies respectively, including the following steps:

[0028] Repeat steps 2)-3) to obtain the distance matrix R between the discrete points and the grid points around the protein through vdw repulsion mat_vdw斥 The distance matrix R of the vdw gravity mat_vdw引 , the fitting results are used as repulsion parameter A and attraction parameter B respectively;

[0029]

[0030]

[0031]

[0032] V ij is the van der Waals interaction energy between two coarse-grained particles, V ij斥 is the van der Waals repulsive interaction energy between two coarse-grained particles, A and B are the parameters to be fitted, r 12 、r 6 The distances between the two particles are respectively the 12th and 6th powers.

[0033] The distance matrix of vdw exclusion is as follows:

[0034]

[0035] It represents the 12th power of the distance between the spatial grid point i around the molecule and the center of the triangular face j;

[0036] The distance matrix of vdw gravity is as follows:

[0037]

[0038] Represents the sixth power of the distance between the spatial grid point i around the molecule and the center of the triangular face j.

[0039] The present invention has the following beneficial effects and advantages:

[0040] According to different systems, appropriate finite element grids can be selected to represent biomacromolecules in triangular grids, and then the electrostatic van der Waals parameters are fitted on the relevant grid surfaces to truly achieve a coarse-grained description of the adjustable scale of biomacromolecules. Its dynamic simulation process is suitable for mesoscopic scale simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a diagram representing biological macromolecules.

[0042] Figure 2 It is a representation of the triangulated mesh of the surface of a biological macromolecule.

[0043] Figure 3 It is a representation of the center point of the finite element surface.

[0044] Figure 4 It is a point charge representation diagram obtained by reproducing and fitting the surrounding electrostatic potential.

[0045] Figure 5 It is the surface electrostatic potential map of biomacromolecules calculated by PBSA. DETAILED DESCRIPTION

[0046] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0047] Cells are the most important research objects in life science and biomedicine, and complex physical and chemical elementary processes such as molecular recognition, enzyme catalysis, and transmembrane material transport occur inside them all the time. Molecular dynamics simulation at the mesoscopic scale is a hot topic today, and the simulation methods that can be used to explore the mesoscopic scale are very limited. Conventional dynamics simulation processes require considerable computing time and very huge computing resources. The content of the present invention is to provide a coarse-grained method for finite element representation of macromolecules. The electrostatic potential and van der Waals potential around the macromolecule are generated by replacing tens of thousands of atoms of irregular macromolecules with a small number of external surfaces (tens or hundreds) of finite elements. The electrostatic and van der Waals parameters of the surface are fitted, and then the molecular dynamics simulation of the finite element mesh is performed using the fitted parameters. We propose the construction of a high-precision finite element model. The present invention is mainly suitable for studying the dynamics and mechanisms of molecular recognition and assembly.

[0048] First, we examine the effect of finite element mesh size on simulation time. Generally speaking, in finite element dynamics simulations, large meshes can be used with large time steps - the increase of both will significantly increase the simulation time scale. To this end, we first examine the step size of the system simulation under different mesh sizes, and then estimate the speed of the system simulation. Taking the simulation of complex ribosomes as an example to illustrate the effect of mesh size on simulation time, the all-atom simulation system contains millions of atoms, and we can see that the minimum mesh size is Adjust to After that, the simulation time scale was increased by nearly 1000 times. Such a large-scale increase makes it possible to simulate the dynamics of huge systems on the millisecond time scale. Table 1 shows the simulation time of a single-core CPU under different grid scales.

[0049] Table 1

[0050]

[0051]

[0052] The present invention proposes a molecular simulation method based on finite element representation. The following are the specific steps of the method:

[0053] 1. First, the molecular surface triangular mesh is obtained through the biological macromolecule structure.

[0054] 2. Take the center of each triangular surface as a candidate charge location.

[0055] 3. Generate the electrostatic potential around irregular surface proteins using the open source dynamics simulation software Tinker.

[0056] 4. Generate a distance matrix between the selected discrete points and the electrostatic potential around the protein.

[0057] 5. The charge of the point charge on the finite element surface is obtained by finding the least squares solution to the overdetermined equation.

[0058] 6. Decompose the van der Waals interaction into repulsive and attractive forces. Fit the repulsive and attractive parameters separately in a way similar to fitting the electrostatic potential to finally obtain the van der Waals parameters of the finite element surface.

[0059] 7. Use the fitted electrostatic and van der Waals parameters to perform molecular finite element dynamics simulation.

[0060] A molecular simulation method based on finite element representation comprises the following steps:

[0061] Step 1. Get the point cloud data of the molecule.

[0062] If the positions of the atoms of the irregular surface protein in the rectangular coordinate system are i=1,2,…,N atom_all .

[0063] The spatial point cloud data of biomacromolecules are obtained by using the coordinates and van der Waals radius of biomacromolecules. For all atoms, the atomic nucleus is taken as the center of the circle and the van der Waals radius is taken as the radius of the sphere. Then, randomly selected points are taken inside each atom to obtain the van der Waals surface morphology of all atoms (including the internal atoms) of the irregular biomacromolecule, forming spatial point cloud data. The coordinates of the atoms of biomacromolecules in the rectangular coordinate system are [x i ,y i ,z i ], the van der Waals radius is The coordinates of a random point in the van der Waals sphere are P i_j , the distance from a random point in the sphere to the nucleus is r ij

[0064] P i_j =rand(x i_j ,y i_j ,z i_j )

[0065] r ij =((x i_j -x i ) 2 +(y i_j -y i ) 2 +(z i_j -z i ) 2 ) 1 / 2

[0066] Where i = 1, 2, ..., N atom_all .

[0067] Step 2. Then do surface triangulation for the point cloud data. The specific algorithm for triangulation is MarchingCubes (the specific algorithm and code can be found in http: / / paulbourke.net / geometry / polygonise / ). This step Generate the original mesh surface file ori.stl .

[0068] Step 3. Convert the fine mesh obtained in the previous step into a coarse mesh according to the needs of the simulation. Set a threshold of the shortest edge length len_min, traverse the mesh edge lengths, and if a certain edge length len_i is less than the threshold len_min, delete the two triangles corresponding to len_i, and move the two endpoints of len_i to the midpoint of len_i. This step generates a coarsened mesh surface file coarse.stl.

[0069] Step 4. Use the coarse mesh surface coarse.stl from the previous step and the open source software tetgen to generate tetrahedral finite elements. The specific commands are as follows:

[0070] tetgen-Yq1.2 / 20V coarse.stl. A brief explanation of the command line options is as follows:

[0071] -q Improve grid (change the quality of the grid).

[0072] -Y Keep the input surface grid (do not modify it).

[0073] TetGen enforces two quality constraints on the tetrahedra: a maximum radius-to-edge ratio constraint and a minimum dihedral angle constraint. By default, the values ​​of these two constraints are 2.0 and 5°, respectively. Alternative quality constraints can be specified after the "-q" command. The two constraints are separated by a slash " / ": the first constraint is the maximum value of the radius-to-edge ratio allowed; the second constraint is the minimum value of the dihedral angle allowed for any tetrahedron. For example, "-q1.2" means the maximum radius-to-edge ratio is 1.2; "-q1.2 / 10" means the maximum radius-to-edge ratio is 1.2 and the minimum dihedral angle is 10°; "-q / 7" means the radius-to-edge ratio uses the default value of 2.0, but the dihedral angle is limited to 7°. You can adjust the above values ​​according to the specific situation when you use it.

[0074] Step 5. Take each external surface triangle (P i_1 ,P i_2 ,P i_3 ) as the geometric center of the charge to be fitted. i

[0075] Oi =(P i_1 +P i_2 +P i_3 ) / 3

[0076] When the position of the point charge at the center of the finite element surface is fixed, we need to solve for the magnitude of the charge. The following are the specific steps.

[0077] Step 6. Generate the electrostatic potential around the protein using tinker software.

[0078] For the protein's pdb file, we call tinker to calculate the electrostatic potential around the protein and get

[0079] Pot i ,i=1,2,…,N pot

[0080] Among them, N pot It is much larger than the number of center points of the finite element surface.

[0081] Step 7. Generate a distance matrix between the selected discrete points and the grid points surrounding the protein on the irregular surface.

[0082] For each spatial point Pos around a protein pot_i , whose potential is determined by the center point (O j ) generated by the charge on j ) in Pos pot_1 The superposition of the electric potential on

[0083]

[0084] in, R ji is the jth triangle mesh center point (O j ) and the spatial grid point around the i-th molecule (Pos pot_i ) between the two. i is the electrostatic potential value of the ith simulated point. The solution process of all points is written in the form of a matrix,

[0085]

[0086] Among them, R mat is a distance matrix.

[0087]

[0088] R ij It is the distance between the spatial grid point i around the molecule and the center of the triangular face j.

[0089] What we want is the potential Pot calculated by tinker i The electric potential sPot simulated by discrete point charges i The gap between them is the smallest. Here we use the principle of least squares method.

[0090] Step 8. Perform QR decomposition on the distance matrix.

[0091] Because N pot is much larger than the number of discrete point charges N cg So here, we solve an overdetermined equation. Specifically, we use the QR decomposition method to decompose the matrix R mat , so that in solving The process is easier to achieve.

[0092] R mat Decompose into two matrices The form of the matrix product of .

[0093]

[0094] in, is an orthogonal square matrix.

[0095]

[0096] in, yes The transpose of , I is the identity matrix. is an upper triangular matrix.

[0097]

[0098]

[0099] Step 9. Fit the charge of the point charge

[0100] because It is an upper triangular matrix, so it is easy to get

[0101] Step 10. Fit the van der Waals repulsion and attraction parameters. The van der Waals interaction potential is in the form of LenardJones 12-6, where the repulsion parameter is represented by A, the attraction parameter is represented by B, and r represents the distance between the two particles. 斥 , V 引 All of them can be calculated using all atoms.

[0102]

[0103]

[0104]

[0105] Tinker software was used to generate grid points around the protein and calculate the grid point van der Waals energy (V 斥 , V 引 ). Generate a distance matrix between the selected discrete points and the grid points around the protein and Similar to fitting electrostatic potential, the difference is to construct the corresponding distance matrix. For each spatial point Pos around the protein pot_i , whose vdw repulsive potential is determined by the discrete points on the protein surface at Pos pot_i The superposition of the vdw repulsive potential.

[0106]

[0107] in Represents the 12th power of the distance between the spatial grid point i around the molecule and the center of the triangular face j.

[0108] Similarly, the distance matrix of vdw gravity can be obtained.

[0109]

[0110] in Represents the sixth power of the distance between the spatial grid point i around the molecule and the center of the triangular face j.

[0111] Step 11. Use the fitted parameters (electrostatic parameters q j , van der Waals repulsion parameters and gravitational parameters) were used for finite element dynamics simulation.

[0112] Before simulation, we must first obtain the molecular surface point cloud data. Then the surface is meshed. The interaction between the center points of the finite element surface can be realized through the electrostatic and van der Waals parameters fitted above. After being subjected to force, it will be distributed to the three vertices on the surface. The constitutive model is used inside the finite element to describe the relationship between stress and strain. The system can describe the time evolution process of N interacting viscoelastic bodies under thermal fluctuations. In viscoelastic bodies, the mechanical response of stress depends not only on the strain, but also on the strain rate. We model viscoelasticity through the Kelvin-Voigt constitutive model. This is one of the simplest models of stress response of viscoelastic solids. Its equation of motion (Cauchy momentum equation) is:

[0113]

[0114] Where ρ represents density, u represents velocity, σ vrepresents the adhesive force, σ e represents elastic force, π represents random force, and f represents external force density.

[0115] like Figure 1 As shown, the biological macromolecule representation diagram; Figure 2 As shown in the figure, the triangulated mesh representation of the biomacromolecule surface. Figure 3 As shown in the figure, the finite element surface center point representation diagram. Figure 4 As shown in , the point charge representation is obtained by reproducing the fitting of the surrounding electrostatic potential. Figure 5 As shown, the surface electrostatic potential map of biomacromolecules calculated by PBSA.

Claims

1. A molecular simulation method based on finite element representation, It is characterized in that The following steps are involved: 1) Use the coordinates and van der Waals radius of the biomacromolecule to obtain the spatial point cloud data of the biomacromolecule, perform surface triangulation on the point cloud data, and generate finite elements; 2) Take the geometric center of each triangular surface of the finite element outer surface as the position of the charge to be fitted; 3) The charge at the center of the finite element surface is obtained by fitting the electrostatic potential around the finite element macromolecules, and the electrostatic potential around the macromolecular protein is generated by the superposition principle to simulate the original electrostatic potential around the protein; 4) The van der Waals parameters of the finite element surface center are obtained by fitting the van der Waals attraction and repulsion potential energies respectively; based on the position of the finite element center and the van der Waals repulsion and attraction parameters, the van der Waals potential around the biomacromolecule is generated through the superposition principle to simulate the original van der Waals potential around the protein.

2. A molecular simulation method based on finite element representation according to claim 1, It is characterized in that Step 2) is as follows: Take each external surface triangle (P i_1 ,P i_2 ,P i_3 ) as the geometric center of the charge to be fitted. i : O i =(P i_1 +P i_2 +P i_3 ) / 3。 3. A molecular simulation method based on finite element representation according to claim 1, It is characterized in that Step 3) is as follows: (1) Generate the electrostatic potential around the protein using the Tinker software; (2) Generate a distance matrix between the selected discrete points and the grid points around the irregular surface protein: (3) Perform QR decomposition on the distance matrix to obtain the upper triangular matrix Fitting the charge of a point charge 4. A molecular simulation method based on finite element representation according to claim 3, It is characterized in that The method of generating a distance matrix between the selected discrete points and the grid points around the irregular surface protein comprises the following steps: For each spatial point Pos around a protein pot_i , whose potential is determined by the discrete point charges on the protein surface at Pos pot_i The superposition of the upper potential, that is in, R ij is the distance between the spatial grid point i around the molecule and the center of the triangular face j, spot i is the electrostatic potential value of the ith simulated value, q j is the charge at the center j of the triangle, N cg is the total number of triangle center points; Write the solution process of all points in the form of a matrix, Among them, R mat is the distance matrix; N cg Expressed as the total number of triangle center points.

5. The molecular simulation method based on finite element representation according to claim 1, It is characterized in that In step 4), the van der Waals parameters of the finite element surface center are obtained by fitting the van der Waals attraction and repulsion potential energies respectively, including the following steps: Repeat steps 2)-3) to obtain the distance matrix R between the discrete points and the grid points around the protein through vdw repulsion mat_vdw斥 The distance matrix R of the vdw gravity mat_vdw引 , the fitting results are used as repulsive force parameter A and attractive force parameter B respectively; V ij is the van der Waals interaction energy between two coarse-grained particles, V ij斥 is the van der Waals repulsive interaction energy between two coarse-grained particles, A and B are the parameters to be fitted, r 12 、r 6 The distances between the two particles are respectively the 12th and 6th powers.

6. A molecular simulation method based on finite element representation according to claim 5, It is characterized in that The distance matrix of vdw exclusion is as follows: It represents the 12th power of the distance between the spatial grid point i around the molecule and the center of the triangular face j; The distance matrix of vdw gravity is as follows: Represents the sixth power of the distance between the spatial grid point i around the molecule and the center of the triangular face j.