Synthetic biology functional element mining method and system, electronic equipment and medium

Abstract

The invention provides a synthetic biology functional element mining method and system, electronic equipment and a medium, and belongs to the technical field of biological element mining, the method comprises the following steps: collecting multi-source heterogeneous data to construct a database, and using a GCNv2 algorithm to extract feature vectors; and optimizing a support vector machine by combining an improved particle swarm optimization algorithm, and constructing a biological functional element classification model. According to user requirements, the model searches for matched candidate biological function elements in a database, and an optimal solution is recommended through a gradient boosting tree sorting algorithm. By means of the synthetic biology functional element mining method and system, the electronic equipment and the medium, the efficiency and accuracy of synthetic biology functional element mining are improved.

Description

Technical Field

[0001] The present invention relates to the field of biological element mining technology, and in particular to a method, system, electronic equipment and medium for mining synthetic biological functional elements. Background Art

[0002] In the field of synthetic biology, the mining of functional elements is the basis for studying and designing biological systems. However, traditional methods for mining functional elements often face many challenges. On the one hand, biological data is multi-source and heterogeneous, including various types of data such as gene sequences, protein structures, and biological function descriptions. These data differ in format, scale, and complexity, which makes data integration and processing difficult. On the other hand, the functions of biological elements are often closely related to their complex structures and interactions. How to effectively extract and use these features for functional classification and mining is an urgent problem to be solved.

[0003] Most existing functional element mining methods rely on experimental verification and manual analysis, which is not only time-consuming and laborious, but also difficult to fully cover all possible biological elements. With the rapid development of bioinformatics, a large amount of biological data is continuously generated and accumulated, providing new opportunities for the mining of functional elements. However, how to efficiently process these multi-source heterogeneous data and mine biological elements with specific functions from them is still an important challenge facing the current field of synthetic biology.

[0004] In addition, traditional classification algorithms have certain limitations in the classification of biological component functions. Due to the complexity and diversity of biological data, traditional classification algorithms often fail to achieve ideal classification results. Therefore, new algorithms and technologies need to be introduced to improve the accuracy and efficiency of biological component function classification.

[0005] In response to the above problems, the present invention proposes a method for mining functional elements in synthetic biology, which aims to achieve efficient and accurate mining and recommendation of functional elements by collecting multi-source heterogeneous data, building a comprehensive database, and adopting advanced feature extraction technology and optimized classification algorithms. Summary of the invention

[0006] The purpose of the present invention is to provide a method, system, electronic equipment and medium for mining synthetic biology functional elements, which improve the efficiency and accuracy of mining synthetic biology functional elements.

[0007] To achieve the above object, the present invention provides a method for mining synthetic biology functional elements, comprising the following steps:

[0008] S1. Collect multi-source heterogeneous data, including gene sequences, protein structures, and biological function descriptions, pre-process the multi-source heterogeneous data, and build a database;

[0009] S2, extracting features from the processed data based on the GCNv2 feature extraction algorithm, generating biological function element feature vectors, and dividing the biological function element feature vectors into a training set and a test set;

[0010] S3, introducing an improved particle swarm optimization algorithm to optimize the support vector machine, and constructing and training a biological functional element classification model based on the training set and the test set in step S2;

[0011] S4. According to the specific needs of users, use the trained biological functional element classification model to search and match in the database, mine candidate biological functional elements that meet the requirements, and recommend the optimal solution based on the sorting algorithm.

[0012] Preferably, in step S1, the protein structure includes primary structure, secondary structure, tertiary structure and quaternary structure; the biological function description includes molecular function, biological process and cellular component.

[0013] Preferably, in step S2, feature extraction is performed on the processed data based on the GCNv2 feature extraction algorithm to generate a biological function element feature vector, and the specific operation is as follows:

[0014] S21, input the preprocessed data into the GCNv2 network, and use the convolution layer to extract features from the preprocessed data;

[0015] f (l+1) =σ(W (l) *f (l) +b (l) )(1);

[0016] Among them, f (l) represents the input feature map of the lth layer; W (l) represents the weight matrix of the lth layer; b (l) represents the bias vector of the lth layer; σ represents the ReLU activation function; * represents the convolution operation; f (l+1) Represents the input feature map of the l+1th layer;

[0017] S22, using the output of step S21 (l+1) Generate probability maps and processed feature maps;

[0018] Use the softmax function to convert f (l+1) Convert to a probability plot:

[0019] o=softmax(f (l+1) )(2);

[0020] Among them, o represents the probability map;

[0021] f(l+1) Perform pooling and normalization operations to obtain the processed feature map f;

[0022] S23, upsampling the probability map to match the resolution of the original image;

[0023] output=PixeShuffle(o,r)(3);

[0024] Where r represents the upsampling factor; output represents the probability map after upsampling; PixeShuffle represents the upsampling function;

[0025] S24, performing non-maximum suppression on the upsampled probability map to determine the position of the feature point;

[0026] S25, using the determined feature point positions, performing bilinear sampling from the processed feature map f obtained in step S22 to extract feature vectors corresponding to the feature point positions;

[0027] desc=BilinearSampling(f,KL)(4);

[0028] Where desc represents the feature vector after bilinear sampling; KL represents the position of the feature point;

[0029] S26, converting the feature vector extracted in step S25 into a binary descriptor through a binary network layer;

[0030] desc' = κ(desc)(5);

[0031] Where desc' represents the binary descriptor feature vector; κ represents the binarization function.

[0032] Preferably, in step S3, an improved particle swarm optimization algorithm is introduced to optimize the support vector machine, and the specific operations are as follows:

[0033] S31, initialize the position X of each particle in the particle swarm i and speed V i , where i represents the index of the particle, and the position represents the hyperparameters C and γ of the SVM;

[0034] Position initialization:

[0035] X i =[C i ,γ i ](6);

[0036] Speed ​​initialization:

[0037] V i =[vC i ,vγi ](7);

[0038] Among them, C i represents the hyperparameter C corresponding to the i-th particle; γ i represents the hyperparameter γ corresponding to the i-th particle; vC i represents the velocity component of the ith particle corresponding to the hyperparameter C; vγ i represents the velocity component of the i-th particle corresponding to the hyperparameter γ;

[0039] S32, define the fitness function, f(X i )=Accuracy(SVM(X i )), where Accuracy is the value of X i Classification accuracy of the trained SVM model;

[0040] S33, dynamically adjust the inertia weight to balance the global search and local search capabilities;

[0041]

[0042] Among them, ω represents the inertia weight; ω max Represents the maximum value of inertia weight; ω min Indicates the minimum value of the inertia weight; t indicates the current number of iterations; iter m Indicates the maximum number of iterations;

[0043] Step S34, dynamically adjusting the learning factors c1 and c2 according to the relationship between the historical best position of the particle and the global best position;

[0044]

[0045] Where c represents the learning factor c1 or c2; c max Indicates the maximum value of the learning factor c1 or c2; c min Represents the minimum value of the learning factor c1 or c2;

[0046] Step S35, updating the speed and position of the particle according to the individual optimal position and the global optimal position;

[0047] V i (T+1)=ωV i (T)+c1×rand()×(pbest i -X i (T)+c2×rand()×(gbest-X i (T)))(10);

[0048] X i (T+1)=X i(T)+V i (T+1)(11);

[0049] Among them, V i (T+1) represents the velocity vector of the i-th particle at time T+1; pbest i represents the best historical position of the ith particle; gbest represents the best historical position of the particle group; X i (T+1) represents the position vector at time T+1; V i (T+1) represents the velocity vector of the i-th particle at time T+1;

[0050] Step S36, updating the individual best position and the global best position;

[0051]

[0052] Where f(X) represents the fitness function; neighborhood (Xi) represents the position set of other particles in the neighborhood of the i-th particle;

[0053] Step S37, repeat steps S33-S36 until the maximum number of iterations is reached;

[0054] Step S38: Output the optimal hyperparameter C * and γ * , and train the final SVM model.

[0055] Preferably, in step S4, according to the specific needs of the user, the trained biological function element classification model is used to search and match in the database, to mine candidate biological function elements that meet the requirements, and to recommend the optimal solution based on the ranking algorithm. The specific operations are as follows:

[0056] S41, using the biological function element classification model trained in step S3, combined with the biological function element feature vector extracted in step S2, to search and match in the constructed database;

[0057] S42, searching for biological components that meet user requirements by using association rules;

[0058] S43, after mining out a plurality of candidate biological function elements, using a ranking algorithm based on a gradient boosting tree to rank these candidate biological function elements to obtain the optimal biological function element;

[0059] S44. Conduct experimental verification on the optimal biological functional element to confirm its actual function and performance;

[0060] S45, feeding back the verification results to the biological functional element classification model to update and optimize the model;

[0061] S46. Output the optimized model results to the user.

[0062] Preferably, in step S42, the biological components meeting the user's requirements are searched using association rules, and the specific operations are as follows:

[0063] S421, processing the biological function element feature vector and biological function description extracted in step S2 to obtain a data form suitable for an association rule mining algorithm;

[0064] S422, generating frequent item sets;

[0065] Frequent itemsets = {itemsets|support(itemsets)≥minimum support} (13);

[0066] Among them, the item set represents the combination of the feature vector of the biological function element extracted from the database and the biological function description; the support (item set) represents the proportion of the item set in all records; the minimum support is a user-defined threshold used to determine whether the item set is frequent;

[0067] S423, generating association rules;

[0068]

[0069] The antecedent represents the premise part of the rule, i.e., a combination of one or more biological function element characteristics; the posterior represents the result part of the rule, i.e., the biological function element characteristics associated with the antecedent;

[0070] S424, calculating the confidence and the lift, and obtaining the filtered rules;

[0071]

[0072] Among them, confidence indicates the probability of the subsequent item occurring under the condition that the antecedent item occurs; support (antecedent ∪ subsequent item) indicates the support for the antecedent and subsequent items to appear at the same time; support (antecedent) indicates the support for the antecedent to appear alone;

[0073]

[0074] Among them, lift represents the improvement of the support of the antecedent to the latter relative to random selection; support (later term) represents the support of the latter when it appears alone;

[0075] S425, assigning an initial score to each candidate biological function element, the score being based on the confidence and lift of its association rule;

[0076] Rating = confidence × lift (17);

[0077] S426. Apply the filtered rules to searches based on customer-specific needs to mine candidate biological functional elements.

[0078] Preferably, in step S43, after mining out a plurality of candidate biological function elements, a ranking algorithm based on a gradient boosting tree is used to sort these candidate biological function elements to obtain the optimal biological function element, and the specific operation is as follows:

[0079] S431, using the biological function element feature vector extracted in step S2 as an input feature and the score of the candidate biological function element obtained in step S425 as a target variable to train a gradient boosting tree model;

[0080] S432, using the test set of step S2 to verify the gradient boosting tree model and evaluate the prediction performance of the model;

[0081] S433, applying the trained gradient boosting tree model to the candidate biological function elements screened in step S42, and ranking all candidate biological function elements according to the scores predicted by the model;

[0082] S434, selecting the element with the highest score from the ranked candidate biological function element list as the optimal biological function element;

[0083] S435 . Output the sorted candidate biological function element list and the optimal biological function element to the user.

[0084] The present invention also provides a synthetic biology functional element mining system, comprising:

[0085] The database construction module is used to collect multi-source heterogeneous data, including gene sequences, protein structures, and biological function descriptions, pre-process the multi-source heterogeneous data, and build a database;

[0086] A feature extraction module, used to extract features from the processed data based on the GCNv2 feature extraction algorithm, generate biological function element feature vectors, and divide the biological function element feature vectors into a training set and a test set;

[0087] Model training and optimization module, used to introduce improved particle swarm optimization algorithm to optimize support vector machine, build and train biological functional element classification model based on training set and test set;

[0088] The functional element mining and recommendation module is used to search and match in the database according to the user's specific needs using the trained biological functional element classification model, to mine candidate biological functional elements that meet the requirements, and to recommend the optimal solution based on the sorting algorithm.

[0089] The present invention also provides a computer device, comprising: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned synthetic biology functional element mining method when executing the computer program.

[0090] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned synthetic biology functional element mining method are implemented.

[0091] Therefore, the present invention adopts the above-mentioned method, system, electronic device and medium for mining functional elements of synthetic biology, and the beneficial technical effects are as follows:

[0092] (1) Efficient data processing capabilities:

[0093] The present invention builds a comprehensive database by collecting multi-source heterogeneous data, including gene sequences, protein structures (covering primary, secondary, tertiary and quaternary structures) and biological function descriptions (including molecular functions, biological processes and cellular components), and preprocessing them. This data integration and preprocessing method significantly improves the efficiency and accuracy of data processing.

[0094] (2) Advanced feature extraction technology:

[0095] The GCNv2 feature extraction algorithm is used to extract features from the processed data and generate feature vectors of biological functional elements. The GCNv2 algorithm effectively extracts the key features of biological elements through steps such as convolutional layers, probability map generation, upsampling, and non-maximum suppression, providing strong support for subsequent classification and mining.

[0096] (3) Optimized model training and classification performance:

[0097] An improved particle swarm optimization algorithm is introduced to optimize the support vector machine (SVM) to build and train a classification model for biological functional elements. This optimization algorithm can dynamically adjust the inertia weight and learning factor, balance the global search and local search capabilities, and thus improve the classification accuracy and generalization ability of the SVM model.

[0098] (4) Accurate mining and recommendation functions:

[0099] According to the specific needs of users, the trained biological functional element classification model is used to search and match in the database to mine candidate biological functional elements that meet the requirements. The optimal biological functional elements are further screened through association rule search and gradient boosting tree-based sorting algorithm, and experimentally verified. This precise mining and recommendation function greatly improves the mining efficiency and accuracy of synthetic biological functional elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1 A flowchart of the method for mining functional elements of synthetic biology of the present invention;

[0101] Figure 2 This is the GCNv2 feature extraction flow chart;

[0102] Figure 3 Optimize graph for particle swarm algorithm;

[0103] Figure 4 Flowchart for association rule mining. DETAILED DESCRIPTION

[0104] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0105] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0106] Embodiment 1

[0107] like Figure 1 FIG. 1 is a flow chart of a method for mining a synthetic biology functional element according to the present invention, comprising the following steps:

[0108] S1. Collect multi-source heterogeneous data, including but not limited to gene sequences, protein structures (primary, secondary, tertiary and quaternary structures) and biological function descriptions (molecular functions, biological processes and cellular components) from public databases.

[0109] Preprocess the collected data, including data cleaning (removing invalid or erroneous data entries), format conversion (unifying the data format for easy processing), and standardization (ensuring that the data is compared and analyzed under the same standards)

[0110] S2. Based on the GCNv2 feature extraction algorithm, feature extraction is performed on the processed data to generate biological function element feature vectors, and the biological function element feature vectors are divided into a training set and a test set. The specific operations are as follows (eg Figure 2 shown):

[0111] S21, input the preprocessed data into the GCNv2 network, and use the convolution layer to extract features from the preprocessed data;

[0112] f (l+1) =σ(W (l) *f (l) +b (l) )(1);

[0113] Among them, f (l) represents the input feature map of the lth layer; W (l)represents the weight matrix of the lth layer; b (l) represents the bias vector of the lth layer; σ represents the ReLU activation function; * represents the convolution operation; f (l+1) Represents the input feature map of the l+1th layer;

[0114] S22, using the output of step S21 (l+1) Generate probability maps and processed feature maps;

[0115] Use the softmax function to convert f (l+1) Convert to a probability plot:

[0116] o=softmax(f (l+1) )(2);

[0117] Among them, o represents the probability map;

[0118] f (l+1) Perform pooling and normalization operations to obtain the processed feature map f;

[0119] S23, upsampling the probability map, and the upsampling factor is set to 2 to match the resolution of the original image;

[0120] output=PixeShuffle(o,r)(3);

[0121] Where r represents the upsampling factor; output represents the probability map after upsampling; PixeShuffle represents the upsampling function;

[0122] S24, performing non-maximum suppression on the upsampled probability map to determine the position of the feature point;

[0123] S25, using the determined feature point positions, performing bilinear sampling from the processed feature map f obtained in step S22 to extract feature vectors corresponding to the feature point positions;

[0124] desc=BilinearSampling(f,KL)(4);

[0125] Among them, desc represents the feature vector after bilinear sampling, and KL represents the position of the feature point;

[0126] S26, converting the feature vector extracted in step S25 into a binary descriptor through a binary network layer;

[0127] desc' = κ(desc)(5);

[0128] Where desc' represents the binary descriptor feature vector; κ represents the binarization function.

[0129] S3, introduce improved particle swarm optimization algorithm to optimize support vector machine (such as Figure 3 As shown), based on the training set and test set of step S2, a biological functional element classification model is constructed and trained, and the specific operations are as follows:

[0130] S31, initialize the position X of each particle in the particle swarm i and speed V i , the number of particles is set to 30, where i represents the index of the particle and the position represents the hyperparameters C and γ of SVM;

[0131] Position initialization:

[0132] X i =[C i ,γ i ](6);

[0133] Speed ​​initialization:

[0134] V i =[vC i ,vγ i ](7);

[0135] Among them, C i is the hyperparameter C corresponding to the i-th particle; γ i is the hyperparameter γ corresponding to the i-th particle; vC i represents the velocity component of the ith particle corresponding to the hyperparameter C; vγ i represents the velocity component of the i-th particle corresponding to the hyperparameter γ;

[0136] S32, define the fitness function, f(X i )=Accuracy(SVM(X i )), where Accuracy is the value of X i Classification accuracy of the trained SVM model;

[0137] S33, dynamically adjust the inertia weight, with the initial value set to 0.9, to balance the global search and local search capabilities;

[0138]

[0139] Among them, ω represents the inertia weight; ω max Indicates the maximum value of the inertia weight, set to 0.95; ω min Indicates the minimum value of the inertia weight, which is set to 0.85; t indicates the current number of iterations; iter m Indicates the maximum number of iterations;

[0140] Step S34, dynamically adjusting the learning factors c1 and c2 according to the relationship between the historical best position of the particle and the global best position;

[0141]

[0142] Where c represents the learning factor c1 or c2; c max represents the maximum value of the learning factor c1 or c2, which is set to 2.5; c min Indicates the minimum value of the learning factor c1 or c2, set to 1.5;

[0143] Step S35, updating the speed and position of the particle according to the individual optimal position and the global optimal position;

[0144] V i (T+1)=ωV i (T)+c1×rand()×(pbest i -X i (T)+c2×rand()×(gbest-X i (T)))(10);

[0145] X i (T+1)=X i (T)+V i (T+1)(11);

[0146] Among them, V i (T+1) represents the velocity vector of the i-th particle at time T+1; pbest i represents the best historical position of the ith particle; gbest represents the best historical position of the particle group; X i (T+1) represents the position vector at time T+1; V i (T+1) represents the velocity vector of the i-th particle at time T+1;

[0147] Step S36, updating the individual best position and the global best position;

[0148]

[0149] Where f(X) represents the fitness function; neighborhood (Xi) represents the position set of other particles in the neighborhood of the i-th particle;

[0150] Step S37, repeat steps S33-S36 until the maximum number of iterations (100 times) is reached;

[0151] Step S38: Output the optimal hyperparameter C * and γ * , and train the final SVM model.

[0152] S4. According to the specific needs of users, use the trained biological functional element classification model to search and match in the database, mine candidate biological functional elements that meet the requirements, and recommend the optimal solution based on the sorting algorithm.

[0153] S41, using the biological function element classification model trained in step S3, combined with the biological function element feature vector extracted in step S2, to search and match in the constructed database;

[0154] S42, using association rules to search for biological components that meet user requirements, the specific operation is as follows (such as Figure 4 shown):

[0155] S421, processing the biological function element feature vector and biological function description extracted in step S2 to obtain a data form suitable for an association rule mining algorithm;

[0156] S422, generating frequent item sets;

[0157] Frequent itemsets = {itemsets|support(itemsets)≥minimum support} (13);

[0158] Among them, the item set represents the combination of the feature vector of the biological function element extracted from the database and the biological function description; the support (item set) represents the proportion of the item set in all records; the minimum support is a user-defined threshold used to determine whether the item set is frequent;

[0159] S423, generating association rules;

[0160]

[0161] The antecedent represents the premise part of the rule, i.e., a combination of one or more biological function element characteristics; the posterior represents the result part of the rule, i.e., the biological function element characteristics associated with the antecedent;

[0162] S424, calculating the confidence and the lift, and obtaining the filtered rules;

[0163]

[0164] Among them, confidence indicates the probability of the subsequent item occurring under the condition that the antecedent item occurs; support (antecedent ∪ subsequent item) indicates the support for the antecedent and subsequent items to appear at the same time; support (antecedent) indicates the support for the antecedent to appear alone;

[0165]

[0166] Among them, lift represents the improvement of the support of the antecedent to the latter relative to random selection; support (later term) represents the support of the latter when it appears alone;

[0167] Selection criteria:

[0168] Minimum confidence: A user-defined threshold used to determine whether a rule is strong or not. A rule is considered meaningful only if its confidence is above this threshold.

[0169] Minimum lift: User-defined threshold used to further filter out rules with substantial impact. A lift greater than 1 indicates that the preceding item has a positive impact on the following item, a lift equal to 1 indicates no impact, and a lift less than 1 indicates a negative impact

[0170] Screening process:

[0171] Calculate confidence and lift: For each generated association rule, calculate its confidence and lift.

[0172] Filter rules: Filter out rules that meet the conditions based on the minimum confidence and minimum lift defined by the user.

[0173] Result output: The meaningful rules screened out are used for subsequent analysis or application.

[0174] S425, assigning an initial score to each candidate biological function element, the score being based on the confidence and lift of its association rule;

[0175] Rating = confidence × lift (17);

[0176] S426. Apply the filtered rules to search for biological components that meet the customer's specific needs.

[0177] S43, after mining a plurality of candidate biological function elements, use a ranking algorithm based on a gradient boosting tree to rank these candidate biological function elements to obtain the optimal biological function element, and the specific operation is as follows:

[0178] S431, using the biological function element feature vector extracted in step S2 as an input feature and the score of the candidate biological function element obtained in step S425 as a target variable to train a gradient boosting tree model;

[0179] S432, using the test set of step S2 to verify the gradient boosting tree model and evaluate the prediction performance of the model;

[0180] S433, applying the trained gradient boosting tree model to the candidate biological function elements screened in step S42, and ranking all candidate biological function elements according to the scores predicted by the model;

[0181] S434, selecting the element with the highest score from the ranked candidate biological function element list as the optimal biological function element;

[0182] S435 . Output the sorted candidate biological function element list and the optimal biological function element to the user.

[0183] S44. Conduct experimental verification on the optimal biological functional element to confirm its actual function and performance;

[0184] S45, feeding back the verification results to the biological functional element classification model to update and optimize the model;

[0185] S46. Output the optimized model results to the user.

[0186] Embodiment 2

[0187] A synthetic biology functional element mining system, comprising:

[0188] The database construction module is used to collect multi-source heterogeneous data, including gene sequences, protein structures, and biological function descriptions, pre-process the multi-source heterogeneous data, and build a database;

[0189] A feature extraction module, used to extract features from the processed data based on the GCNv2 feature extraction algorithm, generate biological function element feature vectors, and divide the biological function element feature vectors into a training set and a test set;

[0190] Model training and optimization module, used to introduce improved particle swarm optimization algorithm to optimize support vector machine, build and train biological functional element classification model based on training set and test set;

[0191] The functional element mining and recommendation module is used to search and match in the database according to the user's specific needs using the trained biological functional element classification model, to mine candidate biological functional elements that meet the requirements, and to recommend the optimal solution based on the sorting algorithm.

[0192] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0193] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0194] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0195] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0196] Therefore, the present invention adopts the above-mentioned synthetic biology functional element mining method, system, electronic equipment and medium to improve the efficiency and accuracy of synthetic biology functional element mining.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for mining functional elements of synthetic biology, characterized in that: The following steps are involved: S1. Collect multi-source heterogeneous data, including gene sequences, protein structures, and biological function descriptions, pre-process the multi-source heterogeneous data, and build a database; S2, extracting features from the processed data based on the GCNv2 feature extraction algorithm, generating biological function element feature vectors, and dividing the biological function element feature vectors into a training set and a test set; S3, introducing an improved particle swarm optimization algorithm to optimize the support vector machine, and constructing and training a biological functional element classification model based on the training set and the test set in step S2; S4. According to the specific needs of users, use the trained biological functional element classification model to search and match in the database, dig out candidate biological functional elements that meet the requirements, and recommend the optimal solution based on the sorting algorithm.

2. A method for mining synthetic biology functional elements according to claim 1, characterized in that: In step S1, the protein structure includes primary structure, secondary structure, tertiary structure and quaternary structure; the biological function description includes molecular function, biological process and cellular component.

3. A method for mining synthetic biology functional elements according to claim 2, characterized in that: In step S2, feature extraction is performed on the processed data based on the GCNv2 feature extraction algorithm to generate a biological function element feature vector. The specific operation is as follows: S21, input the preprocessed data into the GCNv2 network, and use the convolution layer to extract features from the preprocessed data; f (l+1) =σ(W (l) *f (l) +b (l) )(1); Among them, f (l) represents the input feature map of the lth layer; W (l) represents the weight matrix of the lth layer; b (l) represents the bias vector of the lth layer; σ represents the ReLU activation function; * represents the convolution operation; f (l+1) Represents the input feature map of the l+1th layer; S22, using the output of step S21 (l+1) Generate probability maps and processed feature maps; Use the softmax function to convert f (l+1) Convert to a probability plot: o=softmax(f (l+1) )(2); Among them, o represents the probability map; f (l+1) Perform pooling and normalization operations to obtain the processed feature map f; S23, upsampling the probability map to match the resolution of the original image; output=PixeShuffle(o,r)(3); Where r represents the upsampling factor; output represents the probability map after upsampling; PixeShuffle represents the upsampling function; S24, performing non-maximum suppression on the upsampled probability map to determine the position of the feature point; S25, using the determined feature point positions, performing bilinear sampling from the processed feature map f obtained in step S22 to extract feature vectors corresponding to the feature point positions; desc=BilinearSampling(f,KL)(4); Where desc represents the feature vector after bilinear sampling; KL represents the position of the feature point; S26, converting the feature vector extracted in step S25 into a binary descriptor through a binary network layer; desc' = κ(desc)(5); Where desc' represents the binary descriptor feature vector; κ represents the binarization function.

4. A method for mining synthetic biology functional elements according to claim 3, characterized in that: In step S3, an improved particle swarm optimization algorithm is introduced to optimize the support vector machine. The specific operations are as follows: S31, initialize the position X of each particle in the particle swarm i and speed V i , where i represents the index of the particle, and the position represents the hyperparameters C and γ of the SVM; Position initialization: X i =[C i ,c i ](6); Speed ​​initialization: V i =[vC i ,vγ i ](7); Among them, C i represents the hyperparameter C corresponding to the i-th particle; γ i represents the hyperparameter γ corresponding to the i-th particle; vC i represents the velocity component of the ith particle corresponding to the hyperparameter C; vγ i represents the velocity component of the i-th particle corresponding to the hyperparameter γ; S32, define the fitness function, f(X i )=Accuracy(SVM(X i )), where Accuracy is the value of X i Classification accuracy of the trained SVM model; S33, dynamically adjust the inertia weight to balance the global search and local search capabilities; Among them, ω represents the inertia weight; ω max Represents the maximum value of inertia weight; ω min Indicates the minimum value of the inertia weight; t indicates the current number of iterations; iter m Indicates the maximum number of iterations; Step S34, dynamically adjusting the learning factors c1 and c2 according to the relationship between the historical best position of the particle and the global best position; Where c represents the learning factor c1 or c2; c max Indicates the maximum value of the learning factor c1 or c2; c min Represents the minimum value of the learning factor c1 or c2; Step S35, updating the speed and position of the particle according to the individual optimal position and the global optimal position; V i (T+1)=ωV i (T)+c1×rand()×(pbest i -X i (T)+c2×rand()×(gbest-X i (T)))(10); X i (T+1)=X i (T)+V i (T+1) (11); Among them, V i (T+1) represents the velocity vector of the i-th particle at time T+1; pbest i represents the best historical position of the i-th particle; gbest represents the best historical position of the particle group; X i (T+1) represents the position vector at time T+1; V i (T+1) represents the velocity vector of the i-th particle at time T+1; Step S36, updating the individual best position and the global best position; Where f(X) represents the fitness function; the neighborhood (Xi) represents the position set of other particles in the neighborhood of the i-th particle; Step S37, repeat steps S33-S36 until the maximum number of iterations is reached; Step S38: Output the optimal hyperparameter C * and γ * , and train the final SVM model.

5. A method for mining synthetic biology functional elements according to claim 4, characterized in that: In step S4, according to the specific needs of the user, the trained biological function element classification model is used to search and match in the database, to mine candidate biological function elements that meet the requirements, and to recommend the optimal solution based on the ranking algorithm. The specific operations are as follows: S41, using the biological function element classification model trained in step S3, combined with the biological function element feature vector extracted in step S2, to search and match in the constructed database; S42, searching for biological components that meet user requirements by using association rules; S43, after mining out a plurality of candidate biological function elements, using a ranking algorithm based on a gradient boosting tree to rank these candidate biological function elements to obtain the optimal biological function element; S44. Conduct experimental verification on the optimal biological functional element to confirm its actual function and performance; S45, feeding back the verification results to the biological functional element classification model to update and optimize the model; S46. Output the optimized model results to the user.

6. A method for mining synthetic biology functional elements according to claim 5, characterized in that: In step S42, the association rules are used to search for biological components that meet the user's needs. The specific operations are as follows: S421, processing the biological function element feature vector and biological function description extracted in step S2 to obtain a data form suitable for an association rule mining algorithm; S422, generating frequent item sets; Frequent itemsets = {itemsets|support(itemsets)≥minimum support} (13); Among them, the item set represents the combination of the feature vector of the biological function element extracted from the database and the biological function description; the support (item set) represents the proportion of the item set in all records; the minimum support is a user-defined threshold used to determine whether the item set is frequent; S423, generating association rules; The antecedent represents the premise part of the rule, i.e., a combination of one or more biological function element characteristics; the posterior represents the result part of the rule, i.e., the biological function element characteristics associated with the antecedent; S424, calculating the confidence and the lift, and obtaining the filtered rules; Among them, confidence indicates the probability of the subsequent item occurring under the condition that the antecedent item occurs; support (antecedent ∪ subsequent item) indicates the support for the antecedent and subsequent items to appear at the same time; support (antecedent) indicates the support for the antecedent to appear alone; Among them, lift represents the improvement of the support of the antecedent to the latter relative to random selection; support (later term) represents the support of the latter when it appears alone; S425, assigning an initial score to each candidate biological function element, the score being based on the confidence and lift of its association rule; Rating = confidence × lift (17); S426. Apply the filtered rules to searches based on customer-specific needs to mine candidate biological functional elements.

7. A method for mining synthetic biology functional elements according to claim 6, characterized in that: In step S43, after mining a plurality of candidate biological function elements, a ranking algorithm based on a gradient boosting tree is used to sort these candidate biological function elements to obtain the optimal biological function element. The specific operation is as follows: S431, using the biological function element feature vector extracted in step S2 as an input feature and the score of the candidate biological function element obtained in step S425 as a target variable to train a gradient boosting tree model; S432, using the test set of step S2 to verify the gradient boosting tree model and evaluate the prediction performance of the model; S433, applying the trained gradient boosting tree model to the candidate biological function elements screened in step S42, and ranking all candidate biological function elements according to the scores predicted by the model; S434, selecting the element with the highest score from the ranked candidate biological function element list as the optimal biological function element; S435 . Output the sorted candidate biological function element list and the optimal biological function element to the user.

8. A synthetic biology functional element mining system, characterized in that: include: The database construction module is used to collect multi-source heterogeneous data, including gene sequences, protein structures, and biological function descriptions, pre-process the multi-source heterogeneous data, and build a database; A feature extraction module, used to extract features from the processed data based on the GCNv2 feature extraction algorithm, generate biological function element feature vectors, and divide the biological function element feature vectors into a training set and a test set; Model training and optimization module, used to introduce improved particle swarm optimization algorithm to optimize support vector machine, build and train biological functional element classification model based on training set and test set; The functional element mining and recommendation module is used to search and match in the database according to the user's specific needs using the trained biological functional element classification model, to mine candidate biological functional elements that meet the requirements, and to recommend the optimal solution based on the sorting algorithm.

9. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the synthetic biology functional element mining method described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for mining synthetic biology functional elements described in any one of claims 1 to 7 are implemented.

Images