阅读说明：本技术 一种基于混合烟花算法的分子构象搜索方法 (Molecular conformation search method based on mixed firework algorithm ) 是由 李瑾 刘伟超 杨佳艳 于 2020-03-19 设计创作，主要内容包括：本发明提供了一种基于混合烟花算法的分子构象搜索方法,其特征在于：包含以下步骤：S1、设置受体分子的对接区域,用对接盒子表示,对接盒子用于存放配体构象；S2、初始化若干个初始烟花,每个烟花代表一个配体构象；将配体构象表示为解向量,设置受体-配体的结合亲和力打分函数为适应度函数；S3、由解向量构建解空间,解空间包含若干个层次；S4、构建烟花算法的算子；S5、结合烟花算法与局部搜索算法构建混合烟花算法,利用混合烟花算法搜索对接盒子内的近似最优配体构象。本发明减少了分子对接中测试复合物上对接的平均用时,提高了分子对接的速度；同时能够找到适应度函数的近似最优值,提高了分子对接的精度。(The invention provides a molecular conformation search method based on a mixed firework algorithm, which is characterized by comprising the following steps of: comprises the following steps: s1, setting a docking area of the receptor molecule, and representing the docking area by a docking box, wherein the docking box is used for storing ligand conformation; s2, initializing a plurality of initial fireworks, wherein each fireworks represents a ligand conformation; expressing the ligand conformation as a solution vector, and setting a receptor-ligand binding affinity scoring function as a fitness function; s3, constructing a solution space by the solution vector, wherein the solution space comprises a plurality of layers; s4, constructing operators of the firework algorithm; s5, a mixed firework algorithm is constructed by combining the firework algorithm and the local search algorithm, and the mixed firework algorithm is used for searching for the approximately optimal ligand conformation in the butt joint box. The invention reduces the average time consumption of the butt joint on the test compound in the molecular butt joint, and improves the speed of the molecular butt joint; meanwhile, an approximate optimal value of the fitness function can be found, and the precision of molecular docking is improved.)

1. A molecular conformation search method based on a mixed firework algorithm is characterized in that: comprises the following steps:

s1, setting a docking area of a receptor molecule, and representing the docking area by a docking box, wherein the docking box is used for storing ligand conformation;

s2, initializing a plurality of initial fireworks, wherein each initial fireworks represents a ligand conformation; expressing the ligand conformation as a solution vector, and setting a receptor-ligand binding affinity scoring function as a fitness function;

s3, constructing a solution space by the solution vectors, wherein the solution space comprises a plurality of layers which respectively represent different explosion ranges of the fireworks;

s4, constructing operators of the firework algorithm;

and S5, constructing a mixed firework algorithm by combining the firework algorithm and the local search algorithm, and searching the approximately optimal ligand conformation in the butt joint box by using the mixed firework algorithm.

2. The molecular conformation search method based on the mixed firework algorithm as claimed in claim 1, wherein: in the step S3, the solution space includes three levels; the first level refers to the range of variation in the central position of the ligand conformation under conditions of fixed orientation of the ligand conformation and twistable bond angle; the second level refers to the range of orientation rotation of the ligand conformation under the condition that the central position of the ligand conformation and the twistable bond angle are fixed; the third level refers to the range of twistable bond rotations of the ligand conformation under conditions of fixed central position and orientation of the ligand conformation.

3. The molecular conformation search method based on the mixed firework algorithm as claimed in claim 2, wherein: the step S4 includes:

constructing an explosion operator and a mutation operator;

the next generation of firework selection strategy: the next generation of fireworks comprises 1 firework individual with the minimum fitness function value in the candidate set and N-1 firework individual selected according to the roulette method, and the selection probability is calculated according to the difference of the firework individual, wherein the difference is judged by the root mean square error of the atomic position; wherein N is an integer greater than or equal to 2; the candidate set includes current generation fireworks, exploding sparks and variant sparks.

4. The molecular conformation search method based on the mixed firework algorithm as claimed in claim 3, wherein:

the explosion operator comprises a firework explosion range and the number of sparks generated by firework explosion;

the method for calculating the explosion range of the fireworks comprises the following steps: sorting the fireworks according to the fitness function value and dividing the fireworks into three equal parts; the first equal part of fireworks comprises fireworks with fitness function values belonging to a first preset range; the fireworks in the second equal part comprise fireworks with fitness function values belonging to a second preset range; the third equal part of fireworks comprises fireworks with fitness function values belonging to a third preset range;

the calculation formula of the number of sparks generated by the firework explosion is as follows:

S_i＝M-r_i

wherein S is_iIndicating the number of sparks resulting from the explosion of fireworks, M indicating the number of sparks per fireworks which explode at most, r_iAnd the sequence number of the fitness function values of the fireworks is represented.

5. The molecular conformation search method based on the mixed firework algorithm as claimed in claim 4, wherein: the step S5 includes:

s51, local search is conducted on the current-generation fireworks, and whether the local optimal value is accepted or not is determined according to the metropolis criterion;

s52, exploding the current fireworks according to the calculation method of the fireworks explosion range and the calculation formula of the number of sparks generated by fireworks explosion to generate explosion sparks;

s53, carrying out local search on the explosion spark, and determining whether to accept the local optimal value according to metropolis criterion;

s54, randomly selecting L individuals from the contemporary fireworks and the explosion sparks to perform mutation according to a mutation operator to obtain mutation sparks;

s55, constructing the next-generation fireworks according to the next-generation fireworks selection strategy;

s56, if the fitness function value of the firework is converged or the maximum iteration number is reached, the search is ended; otherwise, the process proceeds to step S51.

6. The molecular conformation search method based on the mixed firework algorithm as claimed in any one of claims 1 to 5, wherein the local search algorithm comprises at least one of a BFGS algorithm, a DFP algorithm or an L-BFGS algorithm.

Technical Field

The invention relates to the field of computer-aided drug design, mainly relates to molecular docking, and particularly relates to a conformation search method.

Background

Molecular docking is an important technology in the field of computer-aided drug research, and is widely applied to multiple links of new drug research and development, such as early virtual screening in a drug discovery stage, discovery of drug action targets, research of drug potential action mechanisms, prediction of drug metabolism sites and the like. The molecular docking operation is to place the molecules with known three-dimensional structures on the active sites of the target molecules one by one, predict the binding mode and affinity of the two, and select out the ligand with the best affinity of the receptor close to the natural conformation through a scoring function.

The search for conformation in molecular docking is an extremely complicated problem, and molecular conformation is a different structure of a molecule in which the connection mode of each atom is not changed but the position of each atom with respect to the central atom is changed. From the calculation, the molecular docking is an optimization task, the molecular conformation search algorithm has great influence on the search speed and the hit rate of the system, and the currently adopted algorithm has the problems of low prediction precision and long time for molecular docking.

Disclosure of Invention

In order to achieve the purpose, the invention adopts the following technical scheme: a molecular conformation search method based on a mixed firework algorithm comprises the following steps:

s1, setting a docking area of the receptor molecule, and representing the docking area by a docking box, wherein the docking box is used for storing ligand conformation;

s2, initializing a plurality of initial fireworks, wherein each initial fireworks represents a ligand conformation; expressing the ligand conformation as a solution vector, and setting a receptor-ligand binding affinity scoring function as a fitness function;

s3, constructing a solution space by the solution vectors, wherein the solution space comprises a plurality of layers which respectively represent different explosion ranges of the fireworks;

s4, constructing operators of the firework algorithm;

s5, a mixed firework algorithm is constructed by combining the firework algorithm and the local search algorithm, and the mixed firework algorithm is used for searching for the approximately optimal ligand conformation in the butt joint box.

Compared with the prior art, the invention has the following beneficial effects: the method takes the firework algorithm as a global optimizer to quickly locate promising regions in a solution space, and the local search algorithm is integrated into the firework algorithm to perform local fine search, so that the local search capability is enhanced, the average time of butt joint on a test compound in molecular butt joint is reduced, and the molecular butt joint speed is increased; meanwhile, the approximate solution can be positioned to a local minimum value by a local search algorithm, so that an approximate optimal value of the fitness function is found, and the precision of molecular docking is improved.

Drawings

FIG. 1 is a flow chart of the present invention.

Fig. 2 is a logic block diagram of program execution in embodiment 1.

FIG. 3 is a graph of the binding affinity results of example 1.

FIG. 4 is a graph showing the RMSD results of example 1.

Detailed Description

The technical solution of the present invention is further explained with reference to the drawings and the embodiments.

As shown in fig. 1, the invention provides a molecular conformation search method based on a mixed firework algorithm, which comprises the following steps:

s1, setting a docking area of the receptor molecule, and representing the docking area by a docking box, wherein the docking box is used for storing ligand conformation;

s2, initializing a plurality of initial fireworks, wherein each initial fireworks represents a ligand conformation; expressing the ligand conformation as a solution vector, and setting a receptor-ligand binding affinity scoring function as a fitness function;

s3, constructing a solution space by the solution vectors, wherein the solution space comprises a plurality of layers which respectively represent different explosion ranges of the fireworks;

s4, constructing operators of the firework algorithm;

s5, a mixed firework algorithm is constructed by combining the firework algorithm and the local search algorithm, and the mixed firework algorithm is used for searching for the approximately optimal ligand conformation in the butt joint box.

In step S1, the center position and the side length of the docking box are set, the center of the docking box is set as the geometric center of all the atoms in the receptor binding pocket, and the side length of the docking box is determined by the maximum distance between any two atoms in the receptor binding pocket.

In step S2, the molecular conformation search process is represented using the following optimization problem model:

min F(X)

s.t.g(X)≤0

a constraint function g (x) constrains the ligand conformation in the docking box for the protein, and for all solutions not in the docking box, treats them as invalid solutions; the objective function f (x) corresponds to the receptor-ligand binding affinity scoring function, i.e. the fitness function. The solution to the optimization problem is represented asi is 1,2, … … N, wherein,represents the central position of the conformation of the ligand,is the rotation angle of the ligand conformation, i.e. the orientation information of the ligand conformation,has a value range of [ -pi, pi [ -pi [ ]]；Respectively representing each twistable bond angle in the ligand conformation, and the value range is [ -pi, pi]Wherein 1,2, …, b denotes the number of b twistable bonds in the ligand.

In step S3, a solution space is constructed by the solution vector, the solution space comprises three layers which respectively represent three explosion ranges of the fireworks, wherein the first layer refers to the range of the central position change of the ligand conformation under the condition that the orientation of the ligand conformation and the twistable bond angle are fixed; the second level refers to the range of orientation rotation of the ligand conformation under the condition that the central position of the ligand conformation and the twistable bond angle are fixed; the third level refers to the range of twistable bond rotations of the ligand conformation under conditions of fixed central position and orientation of the ligand conformation.

Step S4 includes:

(1) constructing explosion operators

The explosion operator comprises the firework explosion range and the number of sparks generated by the firework explosion.

The firework explosion range calculation method comprises the following steps: the fireworks are sorted according to the fitness function value from small to large and then divided into three equal parts, preferably, the fireworks in the first equal part are the fireworks with smaller fitness function values, namely the fireworks with the fitness function values belonging to a first preset range and the fireworks with high quality are exploded at the third level, the twistable key angles of the fireworks are randomly changed, and the small-range explosion phenomenon of the fireworks is simulated; the fireworks in the second equal part contain the fireworks with fitness function values belonging to a second preset range, the fireworks are exploded in the second level, and the orientation of the fireworks can be randomly changed to simulate the explosion phenomenon; the fireworks of the third equal portion contain the fireworks that the fitness function value belongs to third preset range, explode in first level, change its central point at random, simulation fireworks explosion phenomenon on a large scale. After the fireworks explosion, explosion sparks are generated, each of which represents a ligand conformation.

The calculation formula of the number of sparks generated by fireworks explosion is as follows:

S_i＝M-r_i

wherein S is_iIndicating fireworks X_iThe number of sparks resulting from the explosion, M being the number of sparks per firework which are most exploded, r_iIndicating fireworks X_iThe sequence number of the fitness function value of (1);

(2) constructing mutation operators

L individuals are randomly selected from the current fireworks and the explosion sparks, a plurality of dimensions are randomly selected from each individual, namely solution vectors, values on the dimensions are multiplied by a variable which is subject to Gaussian distribution to obtain the variant spark individuals, and each variant spark represents a ligand conformation.

The ith sparkThe k variable ofThe variation was performed according to the following formula:

wherein Gaussian (1,1) represents a Gaussian distribution random number with the mean and the variance both being 1, and if the variant spark is an infeasible solution, a variant spark is randomly generated in a solution space.

(3) Firework selection strategy

The selection strategy is that the fireworks with the minimum fitness value in the candidate set are deterministically selected to the next generation as the fireworks, and the remaining N-1 fireworks are selected by a roulette method, wherein N is an integer which is greater than or equal to 2.

In the above formula, RMSD_abThe root mean square error of N atom positions between structures a and b is calculated as the difference between two small molecules a and b with similar three-dimensional structures. Wherein the position of the atom i in the structure a isThe position of the atom i in the structure b is

The above formulaMiddle, RMSD_aThe RMSD of all the firework individuals except a in the current firework individual a to the candidate set K is the sum of the RMSDs of the firework individuals a and K, and the difference of the firework individuals a is represented. If the difference is larger, the number of firework individuals similar to the individual is smaller, and the probability that the individual is selected is larger.

Step S5 includes:

s51, local search is conducted on the current fireworks, whether the local optimum value is accepted or not is determined according to the metropolis criterion, and the local search process is a process of promoting local area excellent individuals;

s52, exploding the current fireworks to generate explosion sparks according to a calculation formula of the fireworks explosion range and a calculation method of the number of the generated sparks;

s53, carrying out local search on the explosion spark, and determining whether to accept the local optimal value according to metropolis criterion;

s54, randomly selecting L individuals from the current fireworks and the explosion sparks to perform mutation according to a mutation operator to obtain the mutation sparks;

s55, selecting N individuals to form the next-generation fireworks according to the next-generation fireworks selection strategy;

s56, if the fitness function value of the firework is converged or the maximum iteration number is reached, the search is ended; otherwise, the process proceeds to step S51.

The local search algorithm comprises at least one of a BFGS algorithm, a DFP algorithm, or an L-BFGS algorithm.