CN107609342A - A kind of protein conformation searching method based on the constraint of secondary structure space length - Google Patents

Abstract

A protein conformation search method based on the spatial distance constraints of the secondary structure. Under the basic framework of the genetic algorithm, the spatial length of each secondary structure in the target protein and the spatial distance between the central residues of two adjacent secondary structures are used. The information constitutes the eigenvector as a space constraint condition, so that under the condition of a given energy function, the solution space is searched in a smaller conformation space, and the spatial distance information is added to the selection operator to make up for the inaccuracy of the energy function Therefore, the accuracy of structural modeling is effectively improved. The invention proposes a protein conformation search method based on secondary structure space distance constraints with high sampling efficiency, high prediction accuracy and low calculation cost.

Description

A Protein Conformation Search Method Based on Space Distance Constraints of Secondary Structure

technical field

The invention relates to the fields of biological informatics, artificial intelligence optimization and computer application, and in particular to a protein conformation search method based on the spatial distance constraints of secondary structures.

Background technique

Protein is a biomacromolecule formed by the dehydration condensation of amino acids, which plays a decisive role in human health. Accurately grasping the structure and function of protein is of great significance to disease research and biopharmaceuticals. Currently, there are two main methods for protein structure prediction: experimental methods and theoretical predictions. Experimental methods include X-ray crystallography, nuclear magnetic resonance spectroscopy, and electron microscopy; although these methods can accurately determine the three-dimensional structure of some proteins, it is time-consuming and expensive to determine the structure through experimental methods, and some proteins The structure is simply not accessible by experimental methods. Therefore, using computational methods to predict protein structures has become a hot spot in bioinformatics research. Theoretical prediction methods mainly use computer technology and intelligent optimization algorithms to predict the three-dimensional structure of proteins from the primary sequence of amino acids, thereby effectively saving the cost of prediction and reducing the time of prediction. Therefore, such methods are more widely used than experimental methods. However, due to the complexity of the protein structure itself, the prediction of the three-dimensional protein structure is still a difficult problem to be solved so far.

In the method of predicting protein structure from scratch, evolutionary algorithm is an important method for studying protein molecular conformation optimization, such as genetic algorithm, differential evolution and other algorithms. These algorithms have the advantages of fast convergence speed, simple structure and strong robustness. However, when the protein sequence is relatively long, because the conformation space is too large, if you search according to a specific energy function, due to the inaccuracy of the energy function, it cannot guarantee that the conformation with the lowest energy found is the closest to the natural state structure, so often The correct fold cannot be formed.

Therefore, the existing conformational space search methods are deficient in prediction accuracy and sampling efficiency and need to be improved.

Contents of the invention

In order to overcome the shortcomings of low sampling efficiency and low prediction accuracy in existing protein structure prediction conformation space search methods, the present invention proposes a protein conformation based on secondary structure space distance constraints with high sampling efficiency and high prediction accuracy Search method.

The technical solution adopted by the present invention to solve its technical problems is:

A protein conformation search method based on secondary structure space distance constraints, said method comprising the following steps:

1) given input sequence information;

2) Parameter initialization: set the population size NP, the maximum genetic algebra G _max , determine the crossover probability P _c , the initial population iteration number iteration, the crossover fragment length frag_length, the assembly counter reject_number, the maximum assembly number reject_max, and the secondary structure in prior knowledge The eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } composed of the spatial length and the spatial distance between the central residues of two adjacent secondary structures, where d _m is the length of the mth secondary structure block of the target protein, d _k,k+1 is the spatial distance between the kth secondary structure block and the k+1th secondary structure central residue, the maximum distance constraint range δ, selection probability P _s ;

3) Initialize the population: start NP Monte Carlo trajectories, search iteration times for each trajectory, and generate NP initial individuals;

4) Perform the following operations on each target individual x _i and randomly selected individual x _j , i, j∈(1,...,NP) and j≠i:

4.1) Carry out crossover operation on individuals x _i and x _j according to probability P _c , the process is as follows:

4.1.1) Randomly select the intersection start point begin_position within the allowable range [1, total_residue-frag_length], and calculate the intersection end point end_position=begin_position+frag_length at the same time, where total_residue is the total number of residues;

4.1.2) Perform twist angle exchange at each intersection position ∈ [begin_position, end_position] to generate new individuals x′ _i , x′ _j , that is, cross individuals x′ _i , x′ _j ;

4.2) Carry out the following mutation operation on cross individuals x′ _i , x′ _j , the process is as follows:

4.2.1) Use the fragment assembly technology to search for the spatial conformation of the crossover individual _x'i , and calculate the length of the secondary structure of the crossover individual _x'i after fragment assembly and the spatial distance between the central residues of two adjacent secondary structures , and form a distance vector in is the length of the mth secondary structure block in the cross individual x′ _i , is the spatial distance between the central residue of the kth secondary structure block and the central residue of the k+1th secondary structure block;

4.2.2) According to the formula Calculate the eigenvector of individual x′ _i The Manhattan distance from the eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } in prior knowledge, if similarity_mutation_1≤δ, the individual x″ _i generated by mutation satisfies Secondary structure space distance constraint, go to step 4.2.4), otherwise go to 4.2.3);

4.2.3) The counter reject_number starts counting, if reject_number≤reject_max, execute steps 4.2.1) and 4.2.2) to generate a new individual x″ _i in sequence, until the similarity_mutation_1≤δ is satisfied; otherwise, execute step 4.2.1) to generate a new individual x″ _i ;

4.2.4) In the same way as in steps 4.2.1) and 4.2.2), perform fragment assembly on the individual _{x'j and calculate the corresponding Manhattan distance value similarity_mutation_2, and finally obtain the new individual x"j ;}

4.2.5) According to the formula Calculate the distance vector of the target individual x _i The Manhattan distance from the eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } in the prior knowledge;

5) Select according to the energy and distance similarity of the target individual x _i and the mutant individual x″ _i , x″ _j , select the dominant individual and update the population, the process is as follows:

5.1) According to the Rosetta Score3 function E( _xi ), calculate the energy E( _xi ), E(x″ _i ) and E(x″ _j ) of the target individual _xi and the mutant individual x″ _i , x″ _j respectively;

5.2) Among the target individual x _i and the mutant individual x″ _i , x″ _j , if the energy value of an individual X, X∈{ _xi ,x″ _i ,x″ _j } is smaller than the energy value of the other two individuals , and the corresponding Manhattan distance value is also smaller than that of the other two individuals, the individual is the dominant individual; if an individual X′,X′∈{ _xi ,x″ _i ,x″ _j } has only If the energy value is smaller than that of the other two individuals, the individual is set as the dominant individual according to the selection probability P _s ; similarly, if an individual X″,X″∈{ _xi ,x″ _i ,x″ _j } Only the corresponding Manhattan distance value is smaller than the corresponding Manhattan distance value of the other two individuals, then the individual is set as the dominant individual according to the selection probability P _s ; finally, the dominant individual replaces the target individual and the population is updated;

6) Judging whether the maximum genetic algebra G _max is reached, if the termination condition is met, then output the result, otherwise go to step 4).

The technical idea of the present invention is: under the basic framework of the genetic algorithm, use the spatial length of each secondary structure in the target protein and the spatial distance information between the central residues of two adjacent secondary structures to form a feature vector as a spatial constraint , so that under the condition of a given energy function, the solution space is searched in a small conformation space, and at the same time, the spatial distance information is added to the selection operator to make up for the inaccuracy of the energy function, thereby effectively improving the structure construction. The accuracy of the model.

The beneficial effect of the present invention is manifested in: on the one hand, the space length of the secondary structure and the space distance between the central residues of two adjacent secondary structures form the eigenvector as the space restriction condition, which reduces the conformational search space and reduces the energy The error caused by the inaccuracy of the function greatly improves the prediction accuracy; on the other hand, under the framework of the genetic algorithm, through the information interaction between individuals and the mutation selection operation of the parent individual, the convergence speed is accelerated and the population density is increased. diversity.

Description of drawings

Figure 1 is a basic flowchart of the protein conformation search method based on the spatial distance constraints of the secondary structure.

Fig. 2 is a schematic diagram of the conformation update of the protein 1AIL when the protein conformation search method based on the space distance constraint of the secondary structure is used for structure prediction.

Figure 3 is a three-dimensional structure diagram obtained by predicting the structure of protein 1AIL based on the protein conformation search method based on the spatial distance constraints of the secondary structure.

detailed description

The present invention will be further described below in conjunction with the accompanying drawings.

Referring to Figures 1 to 3, a protein conformation search method based on secondary structure space distance constraints, the method includes the following steps:

1) given input sequence information;

2) Parameter initialization: set the population size NP, the maximum genetic algebra G _max , determine the crossover probability P _c , the initial population iteration number iteration, the crossover fragment length frag_length, the assembly counter reject_number, the maximum assembly number reject_max, and the secondary structure in prior knowledge The eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } composed of the spatial length and the spatial distance between the central residues of two adjacent secondary structures, where d _m is the length of the mth secondary structure block of the target protein, d _k,k+1 is the spatial distance between the kth secondary structure block and the k+1th secondary structure central residue, the maximum distance constraint range δ, selection probability P _s ;

3) Initialize the population: start NP Monte Carlo trajectories, search iteration times for each trajectory, and generate NP initial individuals;

4) Perform the following operations on each target individual x _i and randomly selected individual x _j , i, j∈(1,...,NP) and j≠i:

4.1) Carry out crossover operation on individuals x _i and x _j according to probability P _c , the process is as follows:

4.1.1) Randomly select the intersection start point begin_position within the allowable range [1, total_residue-frag_length], and calculate the intersection end point end_position=begin_position+frag_length at the same time, where total_residue is the total number of residues;

4.1.2) Perform twist angle exchange at each intersection position ∈ [begin_position, end_position] to generate new individuals x′ _i , x′ _j , that is, cross individuals x′ _i , x′ _j ;

4.2) Carry out the following mutation operation on cross individuals x′ _i , x′ _j , the process is as follows:

4.2.1) Use the fragment assembly technology to search for the spatial conformation of the crossover individual _x'i , and calculate the length of the secondary structure of the crossover individual _x'i after fragment assembly and the spatial distance between the central residues of two adjacent secondary structures , and form a distance vector in is the length of the mth secondary structure block in the cross individual x′ _i , is the spatial distance between the central residue of the kth secondary structure block and the central residue of the k+1th secondary structure block;

4.2.2) According to the formula Calculate the eigenvector of individual x′ _i The Manhattan distance from the eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } in prior knowledge, if similarity_mutation_1≤δ, the individual x″ _i generated by mutation satisfies Secondary structure space distance constraint, go to step 4.2.4), otherwise go to 4.2.3);

4.2.3) The counter reject_number starts counting, if reject_number≤reject_max, execute steps 4.2.1) and 4.2.2) to generate a new individual x″ _i in sequence, until the similarity_mutation_1≤δ is satisfied; otherwise, execute step 4.2.1) to generate a new individual x″ _i ;

4.2.4) In the same way as in steps 4.2.1) and 4.2.2), perform fragment assembly on the individual _{x'j and calculate the corresponding Manhattan distance value similarity_mutation_2, and finally obtain the new individual x"j ;}

4.2.5) According to the formula Calculate the distance vector of the target individual x _i The Manhattan distance from the eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } in the prior knowledge;

5) Select according to the energy and distance similarity of the target individual x _i and the mutant individual x″ _i , x″ _j , select the dominant individual and update the population, the process is as follows:

5.1) According to the Rosetta Score3 function E( _xi ), calculate the energy E( _xi ), E(x″ _i ) and E(x″ _j ) of the target individual _xi and the mutant individual x″ _i , x″ _j respectively;

5.2) Among the target individual x _i and the mutant individual x″ _i , x″ _j , if the energy value of an individual X, X∈{ _xi ,x″ _i ,x″ _j } is smaller than the energy value of the other two individuals , and the corresponding Manhattan distance value is also smaller than that of the other two individuals, the individual is the dominant individual; if an individual X′,X′∈{ _xi ,x″ _i ,x″ _j } has only If the energy value is smaller than that of the other two individuals, the individual is set as the dominant individual according to the selection probability P _s ; similarly, if an individual X″,X″∈{ _xi ,x″ _i ,x″ _j } Only the corresponding Manhattan distance value is smaller than the corresponding Manhattan distance value of the other two individuals, then the individual is set as the dominant individual according to the selection probability P _s ; finally, the dominant individual replaces the target individual and the population is updated;

6) Judging whether the maximum genetic algebra G _max is reached, if the termination condition is met, then output the result, otherwise go to step 4).

The α-fold protein 1AIL with a sequence length of 73 in this example is an example, a protein conformation search method based on the space distance constraint of the secondary structure, which includes the following steps:

1) given input sequence information;

2) Parameter initialization: set the population size NP = 200, the maximum genetic algebra G _max = 2000, determine the crossover probability P _c = 0.1, the initial population iteration number iteration = 2000, the crossover fragment length frag_length = 9, the assembly counter reject_number = 0, the maximum The number of assemblies reject_max=100, the eigenvector D={3.81085,33.8066,8.38603,30.3193,6.69076,22.1852, 19.6409, 17.2739, 15.4455, 14.6372, 15.5907, 12.43}, maximum distance constraint range δ=15, selection probability P _s =0.3;

3) Initialize the population: start NP Monte Carlo trajectories, search iteration times for each trajectory, and generate NP initial individuals;

4) Perform the following operations on each target individual x _i and randomly selected individual x _j , i, j∈(1,...,NP) and j≠i:

4.1) Carry out crossover operation on individuals x _i and x _j according to probability P _c , the process is as follows:

4.1.1) Randomly select the intersection start point begin_position within the allowable range [1, total_residue-frag_length], and calculate the intersection end point end_position=begin_position+frag_length at the same time, where total_residue is the total number of residues;

4.1.2) Perform twist angle exchange at each intersection position ∈ [begin_position, end_position] to generate new individuals x′ _i , x′ _j , that is, cross individuals x′ _i , x′ _j ;

4.2) Carry out the following mutation operation on crossover individuals x′ _i and x′ _j , the process is as follows:

4.2.1) Use the fragment assembly technology to search for the spatial conformation of the crossover individual _x'i , and calculate the length of the secondary structure of the crossover individual _x'i after fragment assembly and the spatial distance between the central residues of two adjacent secondary structures , and form a distance vector in is the length of the mth secondary structure block in the cross individual x′ _i , is the spatial distance between the central residue of the kth secondary structure block and the central residue of the k+1th secondary structure block;

4.2.2) According to the formula Calculate the eigenvector of individual x′ _i The Manhattan distance from the eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } in prior knowledge, if similarity_mutation_1≤δ, the individual x″ _i generated by mutation satisfies Secondary structure space distance constraint, go to step 4.2.4), otherwise go to 4.2.3);

4.2.3) The counter reject_number starts counting, if reject_number≤reject_max, execute steps 4.2.1) and 4.2.2) to generate a new individual x″ _i in sequence, until the similarity_mutation_1≤δ is satisfied; otherwise, execute step 4.2.1) to generate a new individual x″ _i ;

4.2.4) In the same way as in steps 4.2.1) and 4.2.2), perform fragment assembly on the individual _{x'j and calculate the corresponding Manhattan distance value similarity_mutation_2, and finally obtain the new individual x"j ;}

4.2.5) According to the formula Calculate the distance vector of the target individual x _i The Manhattan distance from the eigenvector D={d ₁ ,…,d _m ,d _1,2 ,…,d _k,k+1 } in the prior knowledge;

5) Select according to the energy and distance similarity of the target individual x _i and the mutant individual x″ _i , x″ _j , select the dominant individual and update the population, the process is as follows:

5.1) According to the Rosetta Score3 function E( _xi ), calculate the energy E( _xi ), E(x″ _i ) and E(x″ _j ) of the target individual _xi and the mutant individual x″ _i , x″ _j respectively;

5.2) Among the target individual x _i and the mutant individual x″ _i , x″ _j , if the energy value of an individual X, X∈{ _xi ,x″ _i ,x″ _j } is smaller than the energy value of the other two individuals , and the corresponding Manhattan distance value is also smaller than that of the other two individuals, the individual is the dominant individual; if an individual X′,X′∈{ _xi ,x″ _i ,x″ _j } has only If the energy value is smaller than that of the other two individuals, the individual is set as the dominant individual according to the selection probability P _s ; similarly, if an individual X″,X″∈{ _xi ,x″ _i ,x″ _j } Only the corresponding Manhattan distance value is smaller than the corresponding Manhattan distance value of the other two individuals, then the individual is set as the dominant individual according to the selection probability P _s ; finally, the dominant individual replaces the target individual and the population is updated;

6) Judging whether the maximum genetic algebra G _max is reached, if the termination condition is met, then output the result, otherwise go to step 4).

Taking the α-fold protein 1AIL with a sequence length of 73 as an example, the near-native conformation of the protein was obtained by using the above method, and the minimum root mean square deviation is The average root mean square deviation is The prediction structure is shown in Figure 3.

The above description is the optimization effect obtained by taking 1AIL protein as an example in the present invention, and does not limit the implementation scope of the present invention. Various deformations and improvements are made to it without departing from the scope involved in the basic content of the present invention, and should not be excluded. outside the protection scope of the present invention.

Claims (1)

1. A kind of 1. protein conformation searching method based on the constraint of secondary structure space length, it is characterised in that：The conformation is empty Between searching method comprise the following steps：

   1) list entries information is given；

   2) parameter initialization：Population scale NP, maximum genetic algebra G are set_max, determine crossover probability P_c, initial population iteration time Number iteration, intersects fragment length frag_length, assembles counter reject_number, maximum assembling number Reject_max, space in priori between the space length of secondary structure and two neighboring secondary structure center residue away from From the characteristic vector D={ d of composition₁,…,d_m,d_1,2,…,d_k,k+1, wherein d_mIt is m-th of secondary structure block of target protein Length, d_k,k+1It is the space length of+1 secondary structure center residue of k-th of secondary structure block and kth, ultimate range constrains model Enclose δ, select probability P_s；

   3) population is initialized：Start NP bar Monte Carlo tracks, every track search iteration time, that is, generate NP it is individual at the beginning of Begin individual；

   4) to each target individual x_iWith the individual x randomly selected_jProceed as follows, i, j ∈ (1 ..., NP) and j ≠ i：

   4.1) probability P is pressed_cTo individual x_iAnd x_jCrossover operation is carried out, process is as follows：

   4.1.1) random selection intersects starting point begin_ in allowed band [1, total_residue-frag_length] Position, while cross termination point end_position=begin_position+frag_length is calculated, wherein Total_residue is total number of residues；

   4.1.2) windup-degree is carried out at each intersection site position ∈ [begin_position, end_position] place Exchange, generation new individual x '_i,x′_j, that is, intersect individual x '_i,x′_j；

   4.2) to intersecting individual x '_i,x′_jFollowing mutation operation is carried out, process is as follows：

   4.2.1) using fragment package technique to intersecting individual x '_iSpace conformation search is carried out, calculates and intersects individual x '_iFragment Space length between the length of secondary structure after assembling and two neighboring secondary structure center residue, and form distance vectorWhereinIt is to intersect individual x '_iIn m-th of secondary structure block length,It is The space length of+1 secondary structure block center residue of k-th of secondary structure block center residue and kth；

   4.2.2) according to formulaCalculate individual x '_i Characteristic vectorWith the characteristic vector D={ d in priori₁,…,d_m,d_1,2,…, d_k,k+1Manhattan distances, the individual x " for the generation that made a variation if similarity_mutation_1≤δ_iMeet two level knot Conformational space distance restraint, goes to step 4.2.4), otherwise go to 4.2.3)；

   4.2.3) counter reject_number is started counting up, and is performed successively if reject_number≤reject_max Step 4.2.1) and 4.2.2) generate new individual x "_i, until meeting that similarity_mutation_1≤δ stops；Otherwise perform Step 4.2.1) generation new individual x "_i；

   4.2.4) with step 4.2.1) and 4.2.2) similarly to individual x '_jCarry out fragment assembling and calculate corresponding Manhattan away from From value similarity_mutation_2, new individual x " is finally obtained_j；

   4.2.5) according to formulaCalculate target individual x_i Distance vectorWith the characteristic vector D={ d in priori₁,…,d_m,d_1,2,…, d_k,k+1Manhattan distances；

   5) according to target individual x_iWith the individual x " that makes a variation_i、x″_jEnergy and Distance conformability degree selected, the advantage individual of selecting is simultaneously Population Regeneration, process are as follows：

   5.1) according to Rosetta Score3 function E (x_i) target individual x is calculated respectively_iWith the individual x " that makes a variation_i、x″_jENERGY E (x_i)、E(x″_i) and E (x "_j)；

   5.2) in target individual x_iWith the individual x " that makes a variation_i、x″_jIn, if a certain individual X, X ∈ { x_i,x″_i,x″_jEnergy value be less than Other two individual energy values, while corresponding Manhattan distance values are also than Manhattan corresponding to other two individuals Distance value is small, then the individual is advantage individual；If a certain individual X ', X ' ∈ { x_i,x″_i,x″_jThere was only energy value than other two The energy value of individual is small, then by select probability P_sThe individual is set to advantage individual；Similarly, if a certain individual X ", X " ∈ { x_i, x″_i,x″_jOnly have corresponding Manhattan distance values smaller than Manhattan distance values corresponding to other two individuals, then by choosing Select probability P_sThe individual is set to advantage individual；Finally, advantage individual substitutes target individual, Population Regeneration；

   6) judge whether to reach maximum genetic algebra G_maxIf meeting end condition, output result, step 4) is otherwise gone to.