Seed group updating method, device, equipment, medium and product

Abstract

The invention relates to the technical field of biological information, and discloses a species group updating method and device, equipment, a medium and a product, and the method comprises the steps: obtaining a first target number of initial Boolean networks; encoding the initial Boolean networks of the first target number to obtain initial individuals corresponding to the initial Boolean networks; the initial individual comprises at least one expression tree; constructing an initial population containing initial individuals; acquiring fitness values of the initial individual under one or more specified dimensions; on the basis of the fitness value, updating the expression tree in the initial individual to obtain an updated initial population; determining a target individual based on the updated initial population, and outputting a target Boolean network corresponding to the target individual; the target Boolean network is used for representing the updated population gene. The method can improve the reasoning capability of the Boolean network.

Description

Technical Field

[0001] The present invention relates to the field of bioinformatics technology, and particularly to a population update method, device, equipment, medium and product. Background Art

[0002] With the continuous in-depth study of system biology towards complex gene regulatory networks, traditional continuous mathematical models are facing the bottlenecks of high parameter sensitivity and exponential growth of computational complexity. Due to its discretized logical framework and low parameter dependence characteristics, Boolean networks have been widely applied to large-scale biological systems.

[0003] In related technologies, in the reconstruction of Boolean networks, the population update method usually only targets the topological structure of the Boolean network and ignores the optimization of its internal logical operations, resulting in the situation that the reconstructed Boolean network is prone to correct connections but incorrect logic, affecting the reasoning ability of the Boolean network.

[0004] In view of this, a population update method that can improve the reasoning ability of Boolean networks is needed. Summary of the Invention

[0005] In view of this, the present invention provides a population update method to improve the reasoning ability of Boolean networks.

[0006] In a first aspect, the present invention provides a population update method, which includes: obtaining an initial Boolean network with a first target quantity; respectively encoding the initial Boolean network with the first target quantity to obtain an initial individual corresponding to the initial Boolean network; the initial individual includes at least one expression tree; constructing an initial population including the initial individual; obtaining the fitness value of the initial individual in one or more specified dimensions; based on the fitness value, updating the expression tree in the initial individual to obtain an updated initial population; based on the updated initial population, determining a target individual, and outputting a target Boolean network corresponding to the target individual; the target Boolean network is used to represent the updated population gene.

[0007] In this embodiment, an initial population can be obtained by acquiring a first target number of initial Boolean networks and encoding each of them to obtain initial individuals containing at least one expression tree. By calculating the fitness values of each initial individual in the initial population under one or more specified dimensions, and based on the fitness values, updating the expression trees in the initial individuals to obtain an updated initial population, then determining target individuals therefrom, and outputting the target Boolean networks corresponding to the target individuals. In the above solution, by encoding the initial Boolean networks, the corresponding initial individuals are obtained, and then according to the fitness values of the initial individuals, the expression trees in the initial individuals are updated to obtain an updated initial population. Finally, target individuals are determined from the updated initial population. Compared with encoding Boolean networks using binary strings and other methods, using expression trees can improve the structural interpretability of Boolean networks and the accuracy of updating initial individuals. Thus, the inference ability of Boolean networks is improved.

[0008] In an alternative embodiment, encoding the first target number of initial Boolean networks to obtain the initial individuals corresponding to the initial Boolean networks includes: acquiring the Boolean functions corresponding to each node in the initial Boolean network, and generating expression trees corresponding to the Boolean functions; integrating the expression trees corresponding to each node in the initial Boolean network according to the network topology of the initial Boolean network to generate the initial individuals corresponding to the initial Boolean network.

[0009] In this embodiment, by acquiring the Boolean functions corresponding to each node in the initial Boolean network, generating the corresponding expression trees, and then integrating them to generate the initial individuals corresponding to the initial Boolean network, the structural interpretability of Boolean networks and the global search ability of population update can be improved.

[0010] In an alternative embodiment, the specified dimension includes at least one of the inferred state difference and the average mutual information.

[0011] In this embodiment, the specified dimension includes at least one of the inferred state difference and the average mutual information. When the specified dimension includes the inferred state difference, the speed of population update can be increased and the accuracy of Boolean network inference can be improved; when the specified dimension includes the average mutual information, the diversity of the population can be increased and local optimality can be avoided.

[0012] In an alternative embodiment, updating the expression trees in the initial individuals based on the fitness values includes: using at least one of the update strategies of crossover, mutation, and grafting to update the expression trees of the initial individuals.

[0013] In this embodiment, using at least one of the update strategies of crossover, mutation, and grafting to update the expression trees in the initial individuals can increase the population diversity and avoid local optimality.

[0014] In an alternative embodiment, when the update strategy includes crossover, updating the expression tree in the initial individuals includes: obtaining a first initial individual and a second initial individual, and determining a first expression tree in the first initial individual and a second expression tree in the second initial individual; determining a first node and a second node; the first node is located in the first expression tree, and the second node is located in the second expression tree; swapping the node subtrees corresponding to the first node and the second node respectively to update the initial population.

[0015] In an alternative embodiment, when the update strategy includes mutation, updating the initial population according to a preset update strategy includes: obtaining a third initial individual; randomly selecting a third node; the third node is located in any one of the expression trees in the third initial individual; updating the third node based on a preset mutation strategy and updating the initial population.

[0016] In an alternative embodiment, when the update strategy includes grafting, updating the initial population according to a preset update strategy includes: obtaining a fourth initial individual and a fifth initial individual based on the fitness values of the initial individuals in one or more specified dimensions; the fitness value of the fourth initial individual in one or more specified dimensions is greater than the fitness value of the fifth initial individual in one or more specified dimensions; determining a fourth expression tree in the fourth initial individual and a fifth expression tree in the fifth initial individual; determining a fourth node and a fifth node, the fourth node is located in the fourth expression tree, and the fifth node is located in the fifth expression tree; the prediction error of the Boolean function corresponding to the fourth node is greater than the prediction error of the Boolean function corresponding to the fifth node; replacing the node subtree corresponding to the fourth node with the node subtree corresponding to the fifth node to update the initial population.

[0017] In this embodiment, by selecting a second target number of initial individuals and based on the fitness values of the initial individuals, parent individuals are selected therefrom until a preset number of parent individuals is obtained. The initial population can be optimized and the computational efficiency can be saved.

[0018] In an alternative embodiment, before updating the expression tree in the initial individual based on the fitness value, it includes: selecting a second target number of initial individuals and selecting parent individuals based on the fitness values of the initial individuals; obtaining a preset number of parent individuals; to update the initial population.

[0019] In this embodiment, by obtaining the fitness values of the individuals in the updated initial population in one or more specified dimensions, the target individual is determined. The optimal individual can be screened out to improve the overall performance of the Boolean network.

[0020] Second aspect, the present invention provides a population update device, the device comprising: an initial Boolean network acquisition module for acquiring an initial Boolean network of a first target quantity; an encoding module for respectively encoding the initial Boolean network of the first target quantity to obtain an initial individual corresponding to the initial Boolean network; the initial individual comprising at least one expression tree; a construction module for constructing an initial population comprising the initial individuals; a fitness value acquisition module for acquiring the fitness value of the initial individual under one or more specified dimensions; an update module for updating the expression tree in the initial individual based on the fitness value to obtain an updated initial population; an output module for determining a target individual based on the updated initial population and outputting a target Boolean network corresponding to the target individual; the target Boolean network being used to represent the updated population genes.

[0021] Third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the population update method of the first aspect or any corresponding embodiment thereof.

[0022] Fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to cause a computer to execute the population update method of the first aspect or any corresponding embodiment thereof.

[0023] Fifth aspect, the present invention provides a computer program product comprising computer instructions, the computer instructions being used to cause a computer to execute the population update method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0025] Figure 1 is a flowchart of a population update method according to an embodiment of the present invention;

[0026] Figure 2 is a flowchart of another population update method according to an embodiment of the present invention;

[0027] Figure 3 is a flowchart of another population update method according to an embodiment of the present invention;

[0028] Figure 4 It is a schematic diagram of a Boolean network of a biological model according to an embodiment of the present invention;

[0029] Figure 5 It is a structural block diagram of a population update device according to an embodiment of the present invention;

[0030] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Specific embodiments

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0032] Modeling gene regulatory networks using quantitative and continuous mathematical models has brought important breakthroughs to systems biology. However, these models have problems such as inefficiency in simulating large biological networks and inability to accurately model complex models. Boolean networks can reflect the causal relationships between different parts, predict the transitions of discrete complex systems, and provide an effective and qualitative insight into discrete complex systems.

[0033] In the related art, prior knowledge is usually used to label the Boolean functions in the Boolean network and use them as a chromosome in the genetic algorithm to initialize the population.

[0034] However, using prior knowledge has significant limitations. First, in the case of limited or biased prior knowledge, it will limit the diversity of Boolean functions in the population. Second, it is easily led by prior knowledge into local optima. If the function region pointed to by the prior knowledge does not contain the global optimal solution, then the initialized population may repeatedly search within this local region and it is difficult to jump out of this range. Third, the coding method based on prior knowledge may lead to situations of over-simplification or over-complexity. If the prior knowledge is too simple, then the encoded Boolean function may not be able to fully represent the complex actual situation and lose accuracy. On the contrary, if the prior knowledge itself is very complex, then the encoded chromosome may be too complex to understand and operate, and may introduce too many unnecessary parameters or structures, affecting the efficiency of the genetic algorithm.

[0035] An embodiment of the present invention provides a population update method. By obtaining a first target number of initial Boolean networks and encoding each of them to obtain initial individuals containing at least one expression tree, an initial population can be obtained. By calculating the fitness values of each initial individual in the initial population under one or more specified dimensions, and based on the fitness values, updating the expression trees in the initial individuals to obtain an updated initial population, then determining target individuals therefrom, and outputting the target Boolean networks corresponding to the target individuals. In the above solution, by encoding the initial Boolean networks, the corresponding initial individuals are obtained, and then according to the fitness values of the initial individuals, the expression trees in the initial individuals are updated to obtain an updated initial population, and target individuals are determined therefrom. Compared with encoding Boolean networks using binary strings and other methods, using expression trees can improve the structural interpretability of Boolean networks and the accuracy of updating initial individuals. Thus, the reasoning ability of Boolean networks is improved.

[0036] According to an embodiment of the present invention, an embodiment of a population update method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0037] In this embodiment, a population update method is provided. Figure 1 is a flowchart of the population update method according to an embodiment of the present invention, as Figure 1 shown, the process includes the following steps:

[0038] Step S101, obtain a first target number of initial Boolean networks.

[0039] Since gene regulatory networks are relatively complex, the computational complexity of traditional continuous models is relatively high, and the prediction deviation is large. Using Boolean networks to replace traditional continuous models to model gene regulatory networks can reduce the computational complexity. When using Boolean networks for modeling, in order to improve the performance of the corresponding gene regulatory networks and avoid falling into local optimal traps, a population update method is used for Boolean network reasoning.

[0040] A first target number of initial Boolean networks can be generated based on a generation framework of a random template. Among them, the first target number can be determined according to the network complexity of the Boolean network, that is, in the case where the number of network parameters is relatively large, the first target number is relatively large. The network structure depth of the initial Boolean network does not exceed an input threshold to avoid an overly complex model, where the input threshold can be determined according to the statistical characteristics of the Boolean network.

[0041] Step S102: Encode the initial Boolean networks with the first target quantity respectively to obtain the initial individuals corresponding to the initial Boolean networks; the initial individuals include at least one expression tree.

[0042] An initial Boolean network usually contains multiple nodes, and the state of each node is determined by the Boolean logic combination of its input nodes. Each node in the initial Boolean network can be independently encoded, and each node corresponds to an expression tree, which describes the logical combination of its input signals, so as to directly map the dynamic relationship of multiple nodes and improve the interpretability of the method.

[0043] In some alternative embodiments, a genetic programming approach can also be adopted to automatically generate a computer program corresponding to the Boolean function in the Boolean network. The program structure includes functions and terminal nodes, and each Boolean network is represented by a tree-like structure with a variable length.

[0044] Step S103: Construct an initial population containing the initial individuals.

[0045] Among them, the initial population includes all the initial individuals.

[0046] Step S104: Obtain the fitness values of the initial individuals in one or more specified dimensions.

[0047] The fitness function links the actual problem with the evolutionary mechanism of genetic programming, guides the evolutionary direction by evaluating the performance of individuals, and promotes the population to evolve towards the optimal solution or satisfactory solution of the problem. It plays a core role in genetic programming.

[0048] During the solution process of genetic programming, usually only based on the fitness function, the fitness of each individual in the population is used for searching. The design of the fitness function will directly affect the convergence speed and the final convergence quality of genetic programming. The fitness function can be determined by the staff according to the actual situation.

[0049] The fitness values of the initial individuals in one or more specified dimensions need to be obtained by combining the prediction results of all the expression trees in the individuals. The fitness values can include state matching degree, network simplicity, etc. Among them, the state matching degree characterizes the matching degree between the state transition result predicted by the individual and the actual situation; the network simplicity characterizes the depth of the expression tree, etc., to avoid overfitting.

[0050] When there are multiple specified dimensions, weight values corresponding to the specified dimensions can be set to comprehensively evaluate the initial individuals and avoid the occurrence of local optima.

[0051] Step S105: Update the expression trees in the initial individuals based on the fitness values to obtain an updated initial population.

[0052] Among them, based on the fitness values of each initial individual in one or more specified dimensions, individuals with relatively high fitness in the initial population can be retained first using tournament selection or roulette wheel selection, or individuals with relatively high fitness values in the initial individuals can be directly left, while improving the population update efficiency and avoiding the loss of excellent subtrees.

[0053] After selecting individuals with relatively high fitness, evolutionary strategies such as recombination can be used to update the expression trees in the initial individuals. When the evolutionary strategy includes recombination, two or more initial individuals are selected from the initial population as parent individuals to generate offspring individuals; the nodes of the expression tree in the offspring individuals come from the nodes of the expression tree of one of the parent individuals.

[0054] By updating the expression trees in the initial individuals, the population diversity can be improved.

[0055] Step S106, based on the updated initial population, determine the target individual and output the target Boolean network corresponding to the target individual; the target Boolean network is used to represent the genes of the updated population.

[0056] The individuals included in the updated initial population have changed. The individuals included in the updated population can be evaluated by recalculating the fitness values of the individuals in one or more specified dimensions, and then the target individual can be determined according to the evaluation results. For example, the individual with a large fitness value can be directly selected as the target individual according to the magnitude of the fitness value of the individual.

[0057] After selecting the target individual, convert the expression tree in the target individual into the form of a Boolean network. The target Boolean network is the gene of the updated population.

[0058] The population update method provided in this embodiment can obtain an initial population by obtaining the first target number of initial Boolean networks and encoding them respectively to obtain initial individuals containing at least one expression tree. By calculating the fitness values of each initial individual in the initial population in one or more specified dimensions and updating the expression trees in the initial individuals based on the fitness values to obtain the updated initial population, and then determining the target individual from it and outputting the target Boolean network corresponding to the target individual. In the above solution, by encoding the initial Boolean network, the corresponding initial individual is obtained, and then according to the fitness value of the initial individual, the expression tree in the initial individual is updated to obtain the updated initial population, and the target individual is determined from it. Compared with encoding the Boolean network using binary strings and other methods, using expression trees can improve the structural interpretability of the Boolean network and the accuracy of updating the initial individuals. Thereby improving the reasoning ability of the Boolean network.

[0059] In this embodiment, a population update method is provided.Figure 2 is a flowchart of a population update method according to an embodiment of the present invention. As Figure 2 shown, the process includes the following steps:

[0060] Step S201, obtain an initial Boolean network with a first target quantity.

[0061] For details, please refer to Figure 1 step S101 of the embodiment shown, which will not be elaborated here.

[0062] Step S202, respectively encode the initial Boolean networks with the first target quantity to obtain initial individuals corresponding to the initial Boolean networks; the initial individuals include at least one expression tree.

[0063] Specifically, the above step S202 includes:

[0064] Step S2021, obtain the Boolean functions corresponding to the respective nodes in the initial Boolean network, and generate expression trees corresponding to the Boolean functions.

[0065] The encoding method of the expression tree can be generated by changing the node type and subtree structure, and can represent various forms of Boolean functions, so as to effectively explore the unknown Boolean function space.

[0066] Encoding the initial Boolean network into a tree structure and decomposing the Boolean functions corresponding to the respective nodes into a binary tree form can intuitively reflect the hierarchical relationship of the Boolean functions. And it is convenient to update it subsequently.

[0067] The main parameters of the expression tree include the number of nodes and the depth of the tree. The size of the expression tree is usually proportional to the number of its nodes. The larger the expression tree, the more logical operations and conditions it represents, which may complicate the understanding of the behavior of the Boolean network. Smaller expression trees usually represent simpler logical relationships, making the network easier to understand and interpret.

[0068] Simple logical relationships enable decision-makers to quickly identify the relationship between inputs and outputs. And larger expression trees may lead to overfitting, where the model performs well on the training data but poorly on new data, reducing the interpretability of the model. On the contrary, smaller expression trees usually exhibit better generalization ability, maintaining good performance on unseen data, and also contributing to improving the interpretability of the method.

[0069] In some alternative embodiments, when generating the expression tree corresponding to the Boolean function, restricting the depth of the expression tree can prevent overfitting and improve the efficiency of subsequent population update.

[0070] Step S2022: Integrate the expression trees corresponding to each node in the initial Boolean network according to the network topology of the initial Boolean network to generate an initial individual corresponding to the initial Boolean network.

[0071] According to the network topology of the initial Boolean network, the connection relationship of the expression trees corresponding to each node can be determined. According to the connection relationship, the expression trees corresponding to each node are fused to obtain an initial individual corresponding to the initial Boolean network.

[0072] In some alternative embodiments, when integrating the expression trees corresponding to each node in the initial Boolean network, duplicate sub-expressions can be merged to simplify the logic of the initial individual, thereby improving the efficiency of subsequent searches.

[0073] In some alternative embodiments, in the initial Boolean network G(V, F), which consists of nodes V = {x 1 , x 2 , …, x n , …, x N} and Boolean functions F = {f 1 , f 2 , …, f n , …, f N}, the function expression of any node in the network is:

[0074]

[0075] Wherein, represents the inferred state of the nth node measured at t + 1, represents the Boolean network function of x n deduced by the Boolean function, represents the state of the regulating node of the target node x n at time t.

[0076] When using genetic programming to infer a Boolean network, the Boolean function is encoded into an expression tree: the root node of the expression tree is a logical operation and other operations, and the leaf nodes are Boolean variables, that is, regulating nodes.

[0077] For example, the Boolean expression is where the leaf nodes x a , x b , x c are regulating nodes, and {AND, OR, NOT} represent Boolean operators.

[0078] Encode a Boolean network with a first number of nodes into a first number of groups of expression trees. According to the maximum depth limit of the expression trees, randomly select one from the set of Boolean logics as the root node, and then recursively generate subtrees according to the number of parameters in the Boolean function. By recursively generating subtrees, diverse Boolean expression structures can be constructed, thus improving the population diversity.

[0079] The way of generating subtrees can also be randomly selected from the set of terminals until a pre-set termination condition is met. The termination condition is generally the depth limit for the expression trees. In a specific application, the maximum depth of the expression trees is 4.

[0080] In some alternative embodiments, during the process of encoding a Boolean function into an expression tree, the method of the penalty model in genetic programming can be used to reduce the pressure, thereby restricting the size of the expression tree.

[0081] Step S203, construct an initial population containing initial individuals.

[0082] For details, please refer to Figure 1 Step S103 of the illustrated embodiment, which will not be elaborated here.

[0083] Step S204, obtain the fitness values of the initial individuals in one or more specified dimensions.

[0084] For details, please refer to Figure 1 Step S104 of the illustrated embodiment, which will not be elaborated here.

[0085] In some alternative embodiments, the specified dimensions include at least one of the inferred state difference and the average mutual information.

[0086] Among them, the inferred state difference can be determined according to the difference between the estimated value and the actual value of the nodes in the initial individuals; some nodes in the initial individuals can be selected, the inferred state at a certain moment is obtained, then the actual state at this moment is obtained, and then the difference between the inferred state and the actual state time is calculated to obtain the fitness value in the dimension of the inferred state difference.

[0087] In some alternative embodiments, the root mean square error is used to evaluate the predicted state difference of the nodes in the initial individuals to obtain a quantization value. In a practical application, the difference expression between the inferred state transition and the target node is as follows:

[0088]

[0089] Among them, RMSE(x n ) represents the root mean square error of node x n , and x n(t) represents the inferred state value and the true value of the time series data at time point t; and x n (m) represents the inferred state value and the true value of the time series data in time period m. M = G × T represents the total time length of the time series data, where G represents the number of time series groups and T represents the length of each time series group.

[0090] Among them, node x n The estimated value function of the function expression is as follows:

[0091]

[0092] Among them, is the speculated state of the nth node at time t + 1, is the Boolean network function about x n deduced by the algorithm, is the state of the regulatory node of the target node x n at time t.

[0093] In some alternative embodiments, in order to better solve the overfitting problem and enhance the interpretability of the GP (Gaussian Process) model, the root mean square error and the size of the expression tree can be combined to obtain the prediction state difference of the nodes in the initial individual. In a practical application, the fitness function is as shown in the following formula:

[0094] Fitness 1 (x n ) = -(RMSE(x n ) + λ × SIZE)

[0095] Among them, λ is a regularization parameter used to balance RMSE(x n ) and the expression tree size SIZE.

[0096] The fitness function in the average mutual information dimension can be as shown in the following formula:

[0097]

[0098] Among them, Fitness 2 (x n ) represents the fitness function of node x n in the average mutual information dimension, represents the number of iterations, and AMI represents the average mutual information of the Boolean network. The fitness function value changes with the number of iterations, which can enhance the transparency of the model.

[0099] In some alternative embodiments, the average mutual information is as shown in the following formula:

[0100]

[0101] Among them, K represents the number of nodes in the expression tree, represents the average mutual information between node x n and . represents the k-th node in the expression tree except node x n .

[0102] In some alternative embodiments, the mutual information between two variables is:

[0103] MI(X,Y) = H(X) + H(Y) - H(X,Y)

[0104] where X = (x 1 , x 2 , …, x n ) and Y = (y 1 , y 2 , …, y m ) are discrete random variables, and H(X) and H(Y) respectively represent the information entropy of discrete random variable X and discrete random variable Y. Represents the joint entropy between discrete random variable X and Y. Among them, the information entropy of discrete random variables X and Y and the joint entropy between two random variables are shown in the following formula:

[0105]

[0106] where p(x i ) and p(y i ) respectively represent the marginal probability distributions of discrete random variables X and Y; p(x i , y i ) represents the joint probability distribution of discrete random variables X and Y.

[0107] Step S205: Update the expression tree in the initial individuals based on the fitness values to obtain an updated initial population.

[0108] For details, please refer to Figure 1 Step S105 in the illustrated embodiment, which will not be elaborated here.

[0109] In some alternative embodiments, before the above step S205, the method further includes:

[0110] Step a1: Select a second target number of initial individuals and select parental individuals based on the fitness values of the initial individuals.

[0111] Among them, the second target quantity can be determined according to the actual situation. Each time, the initial individuals of the second target quantity can be selected, and the initial individuals with high fitness values can be selected from them as the parental individuals. In practical applications, to ensure population diversity, the second target quantity is three.

[0112] In some alternative embodiments, after obtaining the fitness values under at least two specified dimensions, non-dominated sorting and crowding distance calculation are respectively performed on them, and the initial individuals with low non-dominated sorting ranks and large crowding distances are selected from the initial individuals of the second target quantity as the parental individuals.

[0113] Step a2, obtain a preset number of parental individuals; to update the initial population.

[0114] By repeating step a1 until a preset number of parental individuals are obtained, the initial population is updated. The initial population can be optimized, making it more targeted when updating the population subsequently, so as to save computational efficiency.

[0115] In some alternative embodiments, the above step S205 includes: adopting at least one update strategy among crossover, mutation, and grafting to update the expression tree of the initial individuals, and obtaining the updated initial population.

[0116] Among them, the corresponding update strategy can be selected according to the actual situation. Among them, crossover means randomly selecting a segment from the expression trees of two individuals for exchange to generate new individuals; mutation means randomly selecting a segment of the expression tree in an individual for random change. Grafting means selecting an excellent expression tree and grafting it into the individual to be improved.

[0117] Specifically, when the update strategy includes crossover, updating the expression tree of the initial individuals includes:

[0118] Step b1, obtain a first initial individual and a second initial individual, and determine the first expression tree in the first initial individual and the second expression tree in the second initial individual.

[0119] Among them, the first initial individual and the second initial individual can be the initial individuals with high fitness values selected from the initial population. Then, the expression trees to be subject to the crossover update strategy are selected from the initial individuals. The first expression tree corresponding to the first initial individual, and the second expression tree corresponding to the second initial individual.

[0120] Step b2, determine a first node and a second node; the first node is located in the first expression tree, and the second node is located in the second expression tree.

[0121] Among them, to improve the effectiveness of the crossover strategy, nodes with high sensitivity can be selected as the first node and the second node based on the sensitivity of each node in the expression.

[0122] Step b3, swap the node subtrees corresponding to the first node and the second node respectively to update the initial population.

[0123] By swapping the node subtrees corresponding to the first node and the second node respectively, two new individuals can be obtained, and the two new individuals are incorporated into the initial population to complete the update of the initial population.

[0124] When the update strategy includes mutation, updating the expression tree of the initial individual includes:

[0125] Step c1, obtain a third initial individual.

[0126] Among them, the one with a high fitness value can be selected from the initial population as the third initial population. The number of the third initial individuals can be determined according to the actual situation.

[0127] Step c2, randomly select a node in at least one expression tree of the third initial individual as the third node.

[0128] Among them, randomly selecting a node in at least one expression tree of the third initial individual as the third node can cover all regions of the expression tree, and there is a chance to mutate the branches with low fitness value but high potential in the expression tree.

[0129] In some alternative embodiments, nodes with high sensitivity can also be selected as the third node according to the sensitivity of each node in the expression tree.

[0130] Step c3, update the third node based on a preset mutation strategy and update the initial population.

[0131] Among them, the mutation strategy can include replacing the subtree corresponding to the third node, changing the operator represented by the third node, etc. By updating the third node, the population diversity can be improved.

[0132] When the update strategy includes grafting, updating the expression tree of the initial individual includes:

[0133] Step d1, obtain a fourth initial individual and a fifth initial individual based on the fitness values of the initial individual in one or more specified dimensions; the fitness value of the fourth initial individual in one or more specified dimensions is greater than the fitness value of the fifth initial individual in one or more specified dimensions.

[0134] Among them, weight assignment can be performed on specified dimensions, and then based on the fitness values of the initial individuals under one or more specified dimensions and the weight assignment corresponding to the specified dimensions, the scores of the initial individuals are determined. Then, according to the scores, the fourth initial individual and the fifth initial individual are determined, where the score of the fourth initial individual is higher than that of the fifth initial individual.

[0135] Step d2: Determine the fourth expression tree in the fourth initial individual and the fifth expression tree in the fifth initial individual.

[0136] Among them, the fourth expression tree and the fifth expression tree can be determined based on the structures of the respective expression trees in the corresponding initial individuals.

[0137] In some alternative embodiments, selecting expression trees with relatively large structural differences for subsequent grafting can reduce the probability of local optimum.

[0138] Step d3: Determine the fourth node and the fifth node. The fourth node is located in the fourth expression tree, and the fifth node is located in the fifth expression tree; the prediction error of the Boolean function corresponding to the fourth node is greater than the prediction error of the Boolean function corresponding to the fifth node.

[0139] Among them, since the prediction error of the Boolean function corresponding to the fourth node is greater than the prediction error of the Boolean function corresponding to the fifth node, in the subsequent grafting, local optimization can be performed on the fourth expression tree to avoid the excellent structure of the fifth node being discarded due to the low fitness value of the fifth initial individual during population update.

[0140] Step d4: Replace the node subtree corresponding to the fourth node with the node subtree corresponding to the fifth node to update the initial population.

[0141] Replacing the node subtree corresponding to the fourth node with the node subtree corresponding to the fifth node can make full use of excellent individuals and increase population diversity.

[0142] Grafting can improve the diversity and adaptability of the population. By combining the advantages of different individuals, the overall performance and evolutionary potential of the population can be significantly improved, thereby enhancing the accuracy and interpretability of the method.

[0143] Step S206: Based on the updated initial population, determine the target individual and output the target Boolean network corresponding to the target individual; the target Boolean network is used to represent the genes of the updated population.

[0144] For details, please refer to Figure 1 Step S106 of the embodiment shown, which will not be elaborated here.

[0145] In some alternative embodiments, the above step S206 includes:

[0146] Step S2061: Obtain the fitness values of a number of first individuals under one or more specified dimensions; the first individuals represent the individuals in the updated initial population.

[0147] Among them, the fitness values of the first individuals under one or more specified dimensions can be referred to in Step S204 and will not be elaborated here.

[0148] Step S2062: Determine the target individual based on the fitness values of a number of first individuals under one or more specified dimensions.

[0149] Based on the fitness values of the first individuals under one or more specified dimensions, select the first individual with a high fitness value as the target individual, and output the Boolean network corresponding to the target individual.

[0150] The population update method provided in this embodiment can improve the structural interpretability of the Boolean network and the global search ability of population update by obtaining the Boolean functions corresponding to each node in the initial Boolean network, generating the corresponding expression trees, and then integrating them to generate the initial individuals corresponding to the initial Boolean network.

[0151] In this embodiment, a population update method is also provided. Figure 3 It is a flowchart of the population update method according to the embodiment of the present invention. As Figure 3 shown, first initialize the population.

[0152] Obtain the expression trees of the individuals in population P, and fuse the scale (SIZE) and root mean square error (RMSE) of the expression trees to obtain the first fitness function (Fitness1). Among them, to balance the scale and the root mean square error, a regularization parameter λ is added to the first fitness function; the root mean square error can be based on time series data; the second fitness function (Fitness2) of the expression tree is determined according to AMI (average mutual information).

[0153] After obtaining the first fitness function value and the second fitness function value of the expression tree, perform non-dominated sorting and crowding distance calculation on them. Select excellent individuals as parent individuals for subsequent operations. In some alternative embodiments, individuals with a low non-dominated sorting rank and a large crowding distance are regarded as excellent individuals.

[0154] In some alternative embodiments, subsequent operations can be performed after generating a sufficient number of parent individuals.

[0155] Using strategies such as mutation, grafting, and crossover to generate a new population Q, then determining whether the termination condition is reached. If the termination condition is reached, the algorithm terminates and outputs the Boolean network. If the termination condition is not reached, calculate the fitness values of the Boolean functions in the new population Q until the termination condition is reached.

[0156] In some alternative embodiments, the termination condition is the number of iterations.

[0157] The population update method provided by the embodiments of the present invention can improve the accuracy of Boolean reasoning by encoding the Boolean network into an expression tree and then using genetic programming for inference. At the same time, during the process of inferring the Boolean network, taking the scale of the expression tree and the search ability of important features as optimization objectives and designing corresponding fitness functions not only enhances the interpretability of the Boolean network model but also solves the overfitting problem commonly existing in the inference process.

[0158] In a practical application, taking an idealized protein - DNA (DeoxyriboNucleic Acid) gene regulatory network as a model, execute as Figure 3 shown in the population update method. Its interaction network is relatively small, containing 6 nodes and the maximum connectivity is 2.

[0159] Mark each node of the gene regulatory network as A, B, C, D, E, and F respectively. Given the states of each node, represented by 0 and 1, where 0 represents inhibition and 1 represents expression; using a Boolean network model to represent the influence relationship between them, it is necessary to find its corresponding mathematical model.

[0160] Here, encode each node of the idealized protein's gene regulatory network (including the state of the node and the logical operation according to the first symbol) into an expression tree in genetic programming, and then initialize the population of genetic programming.

[0161] Then, obtain the fitness function values of nodes A, B, C, D, E, and F in the idealized protein's gene regulatory network. In a practical application, the fitness function values of nodes A, B, C, D, E, and F in the idealized protein's gene regulatory network are shown as follows:

[0162]

[0163] where x 1 , x 2 , x 3 , x 4 , x 5 and x 6 represent nodes A - F respectively. The target node x nThe difference between the inference state and its corresponding value in the time series data T is expressed as RMSE(x n ), SIZE represents the size of the generated expression tree, λ is empirically taken as 0.3, the value of AMI represents the average mutual information value of each node respectively, Fitness 1 , Fitness 2 (x 1 ) are two fitness functions in the process of genetic algorithm inferring Boolean networks.

[0164] Then, new individuals are generated by changing the gene combinations of individuals through crossover, mutation and grafting operations, so as to obtain a new population.

[0165] Among them, in the crossover operation, two individuals randomly select branches to exchange, thus generating new individuals, and the mutation operation is used to randomly change the genes of the new individuals generated by the crossover operation.

[0166] In the mutation operation, the genes of the new individuals generated by the crossover operation are randomly changed.

[0167] In the grafting operation, the chromosomes of excellent individuals are fully utilized and grafted into the individuals to be improved to increase the diversity of the population and explore new solution spaces.

[0168] Repeat the process of calculating fitness values, crossover, mutation and grafting until the number of iterations of population update reaches the preset threshold, then terminate the population update. And determine the target individual from the finally updated population, and output the Boolean network corresponding to the target individual.

[0169] In some alternative embodiments, the preset threshold is set to 200.

[0170] Specifically, the Boolean network topology corresponding to the gene regulatory network of the idealized protein is as Figure 4 shown, and the model of the gene regulatory network is shown as the following formula:

[0171]

[0172] Among them, A(t + 1), B(t + 1), C(t + 1), D(t + 1), E(t + 1) and F(t + 1) respectively represent the states of nodes A to F at time t + 1, and A(t), B(t), C(t), D(t), E(t) and F(t) respectively represent the states of nodes A to F at time t.

[0173] In a practical application, the simulation data of the idealized protein-DNA gene regulatory network is shown in Table 1:

[0174] Table 1. Simulation data

[0175]

[0176] Among them, 1 indicates that the node is connected, 0 indicates that the node is not connected, and the noise ratio is 1%.

[0177] In this embodiment, a population update device is further provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated here. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0178] This embodiment provides a population update device, as Figure 5 shown, including:

[0179] An initial Boolean network acquisition module 501, configured to acquire an initial Boolean network with a first target quantity;

[0180] An encoding module 502, configured to respectively encode the initial Boolean networks with the target quantity to obtain initial individuals corresponding to the initial Boolean networks;

[0181] A construction module 503, configured to construct an initial population including the initial individuals;

[0182] A fitness value acquisition module 504, configured to acquire the fitness values of the initial individuals in one or more specified dimensions;

[0183] An update module 505, configured to update the expression trees in the initial individuals based on the fitness values to obtain an updated initial population;

[0184] An output module 506, configured to determine a target individual based on the updated initial population, and output a target Boolean network corresponding to the target individual; the target Boolean network is used to characterize the genes of the updated population.

[0185] In some alternative implementation manners, the encoding module 502 includes:

[0186] A generation unit, configured to acquire the Boolean functions corresponding to the respective nodes in the initial Boolean network, and generate an expression tree corresponding to the Boolean function;

[0187] An integration unit, configured to integrate the expression trees corresponding to the respective nodes in the initial Boolean network according to the network topology of the initial Boolean network to generate an initial individual corresponding to the initial Boolean network.

[0188] In some alternative implementation manners, the specified dimension includes at least one of an inference state difference and an average mutual information.

[0189] In some alternative embodiments, the update module 505 includes:

[0190] An update unit, configured to update the expression tree of the initial individual by using at least one of the update strategies of crossover, mutation, and grafting.

[0191] In some alternative embodiments, the update unit includes:

[0192] A first individual determination subunit, configured to obtain a first initial individual and a second initial individual, and determine a first expression tree in the first initial individual and a second expression tree in the second initial individual;

[0193] A first target node determination subunit, configured to determine a first node and a second node; the first node is located in the first expression tree, and the second node is located in the second expression tree;

[0194] An exchange subunit, configured to exchange the node subtrees corresponding to the first node and the second node respectively to update the initial population.

[0195] In some alternative embodiments, the update unit further includes:

[0196] A second individual determination subunit, configured to obtain a third initial individual;

[0197] A second target node determination subunit, configured to randomly select a third node; the third node is located in any one of the expression trees in the third initial individual;

[0198] A mutation subunit, configured to update the third node based on a preset mutation strategy and update the initial population.

[0199] In some alternative embodiments, the update unit further includes:

[0200] A third individual determination subunit, configured to obtain a fourth initial individual and a fifth initial individual based on the fitness values of the initial individuals in one or more specified dimensions; the fitness value of the fourth initial individual in one or more specified dimensions is greater than the fitness value of the fifth initial individual in one or more specified dimensions;

[0201] An expression tree determination subunit, configured to determine a fourth expression tree in the fourth initial individual and a fifth expression tree in the fifth initial individual;

[0202] A third target node determination subunit, configured to determine a fourth node and a fifth node, the fourth node is located in the fourth expression tree, and the fifth node is located in the fifth expression tree; the Boolean function prediction error corresponding to the fourth node is greater than the Boolean function prediction error corresponding to the fifth node;

[0203] The grafting subunit is used to replace the node subtree corresponding to the fourth node with the node subtree corresponding to the fifth node to update the initial population.

[0204] In some alternative embodiments, the device further includes a preference module for updating the initial population. The preference module includes:

[0205] The parental individual selection unit is used to select a second target number of initial individuals and select parental individuals based on the fitness values of the initial individuals;

[0206] The preference unit is used to obtain a preset number of parental individuals to update the initial population.

[0207] In some alternative embodiments, the output module 506 includes:

[0208] The fitness calculation unit is used to obtain the fitness values of a number of first individuals in one or more specified dimensions; the first individuals represent the individuals in the updated initial population;

[0209] The target individual determination unit is used to determine the target individual based on the fitness values of a number of first individuals in one or more specified dimensions.

[0210] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding foregoing embodiments, and will not be elaborated herein.

[0211] The population update device in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0212] The embodiment of the present invention further provides a computer device having the above Figure 5 shown population update device.

[0213] Please refer to Figure 6 , Figure 6 which is a schematic structural diagram of a computer device provided by an alternative embodiment of the present invention. As Figure 6As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some alternative embodiments, if needed, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor system). Figure 6 Taking one processor 10 as an example in

[0214] The processor 10 can be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 can further include a hardware chip. The above hardware chip can be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above programmable logic device can be a complex programmable logic device, a field programmable gate array, a general array logic, or any combination thereof.

[0215] Among them, the memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiments.

[0216] The memory 20 can include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 20 can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 can optionally include a memory remotely set relative to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0217] The memory 20 can include a volatile memory, such as a random access memory; the memory can also include a non-volatile memory, such as a flash memory, a hard disk, or a solid-state drive; the memory 20 can also include a combination of the above types of memories.

[0218] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 can be connected through a bus or other means.Figure 6 Take the bus connection as an example.

[0219] The input device 30 can receive input digital or character information, and generate key signal inputs related to the user settings and function controls of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED), and a haptic feedback device (e.g., a vibration motor), etc. The above display device includes but is not limited to a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device may be a touch screen.

[0220] The embodiments of the present invention also provide a computer-readable storage medium. The method according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented by downloading through a network the original computer code stored in a remote storage medium or a non-transitory machine-readable storage medium and to be stored in a local storage medium, so that the method described herein can be stored as such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiments is implemented.

[0221] A part of the present invention can be applied as a computer program product, such as computer program instructions, which when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should be able to understand that the forms of existence of computer program instructions in a computer-readable medium include but are not limited to source files, executable files, installation package files, etc. Correspondingly, the ways for computer program instructions to be executed by a computer include but are not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible by the computer.

[0222] Although embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A population updating method, characterized in that: The method comprises: Obtaining a first target number of initial Boolean networks; Encoding the first target number of initial Boolean networks respectively to obtain initial individuals corresponding to the initial Boolean networks; the initial individuals include at least one expression tree; constructing an initial population including the initial individuals; Obtaining the fitness value of the initial individual in one or more specified dimensions; Based on the fitness value, the expression tree in the initial individual is updated to obtain an updated initial population; Based on the updated initial population, a target individual is determined, and a target Boolean network corresponding to the target individual is output; the target Boolean network is used to characterize the updated population gene.

2. The method according to claim 1, characterized in that The encoding of the initial Boolean network for the first target number to obtain initial individuals corresponding to the initial Boolean network includes: Obtaining the Boolean function corresponding to each node in the initial Boolean network, and generating an expression tree corresponding to the Boolean function; According to the network topology of the initial Boolean network, the expression trees corresponding to the nodes in the initial Boolean network are integrated to generate the initial individuals corresponding to the initial Boolean network.

3. The method according to claim 1, characterized in that: The specified dimension includes at least one of an inferred state difference and an average mutual information.

4. The method according to claim 1, characterized in that The updating of the expression tree in the initial individual based on the fitness value includes: At least one updating strategy among crossover, mutation and grafting is adopted to update the expression tree of the initial individual.

5. The method according to claim 4, characterized in that In the case where the update strategy includes crossover, the updating of the expression tree in the initial individual includes: Obtain a first initial individual and a second initial individual, and determine a first expression tree in the first initial individual and a second expression tree in the second initial individual; Determine a first node and a second node; the first node is located in the first expression tree, and the second node is located in the second expression tree; The node subtrees corresponding to the first node and the second node are exchanged to update the initial population.

6. The method according to claim 4, characterized in that In the case where the update strategy includes mutation, updating the initial population according to a preset update strategy includes: Obtaining a third initial individual; Randomly select a third node; the third node is located in any expression tree of the third initial individual; Based on a preset mutation strategy, the third node is updated, and the initial population is updated.

7. The method according to claim 4, characterized in that In the case where the updating strategy includes grafting, updating the initial population according to the preset updating strategy includes: Based on the fitness value of the initial individual in one or more specified dimensions, obtaining a fourth initial individual and a fifth initial individual; the fitness value of the fourth initial individual in one or more specified dimensions is greater than the fitness value of the fifth initial individual in one or more specified dimensions; determining a fourth expression tree in a fourth initial individual and a fifth expression tree in a fifth initial individual; Determine a fourth node and a fifth node, the fourth node is located in the fourth expression tree, and the fifth node is located in the fifth expression tree; the Boolean function prediction error corresponding to the fourth node is greater than the Boolean function prediction error corresponding to the fifth node; The node subtree corresponding to the fourth node is replaced by the node subtree corresponding to the fifth node, so as to update the initial population.

8. The method according to claim 4, characterized in that Before updating the expression tree in the initial individual based on the fitness value, the method includes: Selecting a second target number of initial individuals, and selecting parent individuals based on the fitness values ​​of the initial individuals; Get a preset number of parent individuals to update the initial population.

9. The method according to claim 1, characterized in that: The step of determining the target individual based on the updated initial population includes: Obtaining fitness values ​​of a plurality of first individuals in one or more specified dimensions; the first individuals represent individuals in the updated initial population; Based on the fitness values ​​of the first individuals in one or more specified dimensions, a target individual is determined.

10. A population updating device, characterized in that: The device comprises: An initial Boolean network acquisition module, used to acquire a first target number of initial Boolean networks; An encoding module, used for encoding the first target number of initial Boolean networks respectively, and obtaining initial individuals corresponding to the initial Boolean networks; the initial individuals include at least one expression tree; A construction module, used for constructing an initial population including the initial individuals; A fitness value acquisition module, used to obtain the fitness value of the initial individual in one or more specified dimensions; An updating module, used for updating the expression tree in the initial individual based on the fitness value to obtain an updated initial population; The output module is used to determine the target individual based on the updated initial population and output a target Boolean network corresponding to the target individual; the target Boolean network is used to characterize the updated population gene.

11. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the population updating method according to any one of claims 1 to 9 by executing the computer instructions.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the population updating method according to any one of claims 1 to 9.

13. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the population updating method according to any one of claims 1 to 9.

Images