A method and apparatus for protein structure classification and detection based on minimizing information structure entropy.

Abstract

This application provides a protein structure classification and detection method and apparatus based on minimizing information structure entropy. The method includes: acquiring protein molecule growth trajectory data and constructing a directed graph network set; initializing the directed graph network set as a first coding tree, calculating the information structure entropy of the leaf nodes of the first coding tree, and generating a second coding tree with height k; calculating the optimal coding tree corresponding to each coding tree in the second coding tree using an information structure entropy minimization algorithm, and constructing an optimal coding tree set; inputting the optimal coding tree set into a pre-constructed graph convolutional neural network model, outputting predicted labels, and obtaining the protein structure classification result based on the predicted labels. The protein structure classification and detection method and apparatus based on minimizing information structure entropy described in this application can accurately and efficiently classify and detect protein structures, and at the same time, can accelerate the prediction of protein structures.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of protein structure classification detection, and particularly relates to a protein structure classification detection method and device based on minimized information structure entropy. BACKGROUND

[0002] Proteins are very important biological macromolecules in living organisms, which play key functional and structural roles in cells, such as catalyzing metabolic reactions, transmitting signals, and supporting cell structures. Due to the diversity and complexity of protein molecular structures, the classification and research of protein molecular structures have become an important research direction in the fields of biology and biochemistry.

[0003] At present, protein structure prediction research mainly focuses on machine learning-based methods, sequence feature-based methods, and protein interaction-based methods. Among them, machine learning-based methods mainly use algorithms such as support vector machines and random forests to analyze and learn protein sequences to predict their structures and interaction patterns. Although these methods can achieve good prediction results in some cases, for some complex protein structures and application scenarios, these methods may obtain too high a dimension when obtaining protein features, resulting in a long experimental period, consumption of a large amount of time and resources, and certain errors and uncertainties in the prediction results. SUMMARY

[0004] Therefore, the present application aims to provide a protein structure classification detection method and device based on minimized information structure entropy to solve the problem of certain errors and uncertainties in the prediction results for some complex protein structures and application scenarios during protein structure classification detection.

[0005] To achieve the above-mentioned purpose, the technical scheme of the present application is as follows:

[0006] The present application provides a protein structure classification detection method based on minimized information structure entropy, which comprises:

[0007] Obtaining growth trajectory data of protein molecules and constructing a directed graph network set;

[0008] Initializing the directed graph network set as a first encoding tree, calculating the information structure entropy of the leaf nodes of the first encoding tree, and generating a second encoding tree with a height of k;

[0009] Calculating each encoding tree in the second encoding tree through an information structure entropy minimization algorithm to obtain an optimal encoding tree corresponding to each encoding tree, and constructing an optimal encoding tree set;

[0010] The optimal coding tree set is input into a pre-constructed graph convolutional neural network model to output a predicted label, and a classification result of the protein structure is obtained according to the predicted label.

[0011] Based on the same inventive concept, the present application also provides a protein structure classification detection device based on minimum information structure entropy, comprising:

[0012] An acquisition module is configured to acquire growth trajectory data of a protein molecule and construct a directed graph network set.

[0013] A coding tree generation module is configured to initialize the directed graph network set as a first coding tree, calculate information structure entropy of leaf nodes of the first coding tree, and generate a second coding tree with a height of k.

[0014] An optimal coding tree construction module is configured to calculate each coding tree in the second coding tree by an information structure entropy minimization algorithm, obtain an optimal coding tree corresponding to each coding tree, and construct an optimal coding tree set.

[0015] A result classification module is configured to input the optimal coding tree set into a pre-constructed graph convolutional neural network model to output a predicted label, and obtain a classification result of the protein structure according to the predicted label.

[0016] Based on the same inventive concept, the present application also provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the protein structure classification detection method based on minimum information structure entropy as described above.

[0017] Based on the same inventive concept, the present application also provides a non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores computer instructions for causing the computer to execute the protein structure classification detection method based on minimum information structure entropy as described above.

[0018] Compared with the prior art, the protein structure classification detection method and device based on minimum information structure entropy have the following beneficial effects:

[0019] The protein structure classification detection method and device based on minimizing information structure entropy provided in the application, the method generates a coding tree by processing the growth trajectory data of the obtained protein molecules, calculates the coding tree by the minimizing information structure entropy algorithm, obtains an optimal coding tree set, uses a graph convolutional neural network model to predict the classification of the protein structure, and obtains the classification result of the protein molecule structure. The method improves the prediction speed of the complex protein molecules, improves the accuracy and efficiency of the protein molecule classification, and has important significance for the research of proteins and the development of drugs. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The embodiments of this application illustrated in the drawings, and their description, are presented to explain the application and are not intended to limit the application. In the drawings:

[0021] Figure 1 The protein structure classification detection method flow chart based on minimizing information structure entropy described in the embodiments of the application;

[0022] Figure 2 The protein structure classification detection device structure schematic diagram based on minimizing information structure entropy described in the embodiments of the application;

[0023] Figure 3 The electronic device structure schematic diagram described in the embodiments of the application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below with reference to the embodiments and the accompanying drawings.

[0025] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the application should be understood as the usual meaning understood by those skilled in the art to which the embodiments of the application belong. The terms "first", "second" and the like used in the embodiments of the application do not represent any order, number or importance, but are only used to distinguish different components. The terms "include" or "contain" and the like mean that the elements or objects before the terms cover the elements or objects listed after the terms and their equivalents, and do not exclude other elements or objects. The terms "connect" or "connected" and the like do not mean physical or mechanical connection, but can include electrical connection, whether direct or indirect. The terms "up", "down", "left", "right" and the like only represent relative positional relationships, which may change when the absolute positions of the described objects change.

[0026] The embodiments of the application are described in detail below with reference to the accompanying drawings.

[0027] Referring to Figure 1 As shown in the figure, the protein structure classification detection method based on minimizing information structure entropy of an embodiment of the present application comprises the following steps:

[0028] S1, obtaining growth trajectory data of protein molecules, and constructing a directed graph network set.

[0029] Specifically, the growth trajectory data of the protein molecules is obtained, and the growth trajectory data is converted into an adjacency matrix.

[0030] A directed graph network is constructed according to the adjacency matrix, wherein the direction of the edge in the directed graph network is determined according to the growth order of the protein molecules.

[0031] The directed graph network generated by each protein molecule is integrated into a directed graph network set.

[0032] Specifically, the growth trajectory data of the protein molecules is read, in which each amino acid is labeled as an atom, a file storing protein structure data is deconstructed, and the data in the file is read using python to convert the data into an adjacency matrix A, wherein A ∈ R n*n , R represents a real number set, and n is the dimension.

[0033] A directed graph network is constructed according to the adjacency matrix A, and the direction of the edge in the graph network is determined according to the growth order of the protein molecules. In the adjacency matrix A, if A ij ≠ 0, it represents that there is a covalent bond between atom i and atom j in the protein, the direction of the directed edge is i pointing to j, and the value of A ij represents the weight of the directed edge.

[0034] The directed graph network generated by each protein is integrated into a new directed graph network set G = {G1, G2, …}.

[0035] S2, initializing the directed graph network set as a first coding tree, calculating the information structure entropy of the leaf nodes of the first coding tree, and generating a second coding tree with a height of k.

[0036] Specifically, the directed graph network set is initialized as a coding tree, the information structure entropy of the leaf nodes of the coding tree is calculated, and a coding tree with a certain height is generated through three operations of inserting, deleting and comparing the height of the leaf nodes.

[0037] In some embodiments, the leaf nodes of the first coding tree are subjected to node merging entropy calculation, node removal entropy calculation and node height comparison to generate a second coding tree with a height of k.

[0038] Specifically, it comprises the following steps:

[0039] S21, according to the directed graph network set G constructed in step S1, and the preset height k of the target coding tree;

[0040] S22, initialize the directed graph network set G={G1, G2, …} as the coding tree T={T1, T2, …};

[0041] wherein, for any coding tree T i , v r is the root node of any coding tree T i , V is the leaf node set of T i ; The attributes of any node v i of T i include parent, children, vol, g, height, parent represents the parent node of node v i , children represents the child nodes of the node, vol represents the in-degree weighted sum of the node, g represents the sum of the vol of the node and the in-degree weighted sum of the child nodes, height represents the height of the node in the coding tree T i ;

[0042] S23, node merging entropy calculation, node removal entropy calculation and node height comparison are performed on the leaf nodes of the first coding tree to generate a second coding tree with height k.

[0043] Further, for any coding tree T i , nodes v r and v i are selected from the leaf nodes of the root node v j using the node merging entropy calculation formula, a new node v e is created, and the node v e is inserted between v r and (v i , v j ); After all the leaf nodes are merged, if the height attribute height of the root node v r is greater than the first threshold value, the leaf node and its parent node are merged again until the height attribute height of the root node v r is equal to the first threshold value; The first threshold value set in this embodiment is 3, and the first threshold value in the following embodiments is also taken as an example. It should be noted that the specific value of the first threshold value can be flexibly set according to actual conditions, which will not be described further here.

[0044] The specific steps are as follows:

[0045] S2311, from the root node v i of the coding tree T rThe node merging entropy of the computing node in the leaf node is calculated, and a node v i and v j The calculation formula of the node merging entropy is:

[0046]

[0047] Wherein, v i is the weighted in-degree sum of the node v i , g i is the sum of the weighted in-degree sum of the node v i and the weighted in-degree sum of the child node, (v i -g i ) is the weighted in-degree sum of the child node of the node v i , v is the sum of the weighted in-degree sum of the node v i and the weighted in-degree sum of the node v j , v j , g j , (v j -g j ) are the same; vol(V) is the sum of the weighted in-degree of the whole tree; represents obtaining (v i , v j ) with the minimum value;

[0048] The implementation method is that, each time, an arbitrary pair of adjacent leaf nodes is taken, the node merging entropy of the pair of nodes is calculated through the node merging entropy calculation formula, and the calculation result is stored in the minimum heap. In the minimum heap, the top element of the heap is the minimum value, and the node (v i , v j ) with the minimum node merging entropy can be obtained by taking out the top element of the heap;

[0049] vol(V) is represented by the following formula:

[0050]

[0051]

[0052] Wherein, w is determined by the weight function E→R + on the edge, represents the sum of the weights of all in-degree edges of the node v;

[0053] S2312, a new node v e is created;

[0054] S2313, the node v e is inserted, and the steps are as follows:

[0055] v i → v e.children

[0056] v j →v e .children

[0057] v e →v r .children

[0058] Insert node v i and node v j into the child nodes of node v e , v e is inserted into the child nodes of root node v r ;

[0059] S2314, update the properties of v e node, the height attribute takes the maximum value in the child node, the formula is as follows:

[0060] g e = g i + g j

[0061] v e .height = Max(v i .height, v j .height) + 1

[0062] S2315, if all leaf nodes are not merged, jump to S2311;

[0063] S2316, detect whether the height attribute height of root node v i in the coding tree T r is greater than 3; if greater than 3, select the parent node v j of the leaf node, move the leaf node to the parent node of node v j , remove node v j ; wherein the value of the height attribute of each node is the maximum value of the height of the child node; the height attribute of the root node v r is updated according to the following formula:

[0064] v r .height = Max(v r .children.height) + 1

[0065] S2317, if the height attribute height of root node v i in the coding tree T r is greater than 3, jump to S2316.

[0066] Further, if the height of the coding tree T i is greater than the input k, a node removal entropy is used to select a node v i from the tree T i , remove the node v i , and compress the height of the tree to k layers by removing the node.

[0067] The specific steps are as follows:

[0068] S2321, detecting whether the height of the coding tree T i is greater than the input height k, if the height of the coding tree T i is greater than the input coding tree height parameter k, then execute downwardly;

[0069] S2322, calculating a node removal entropy in the coding tree T i , and selecting a minimum removal entropy node v i from the T i , removing the same; wherein the node v i cannot be a root node or a leaf node, and the calculation formula of the node removal entropy is:

[0070]

[0071] wherein H T (REMOVE(v i )) is the structure entropy of the coding tree after deleting the node v i , and REMOVE(·) is a node removal function;

[0072] S2323, the REMOVE(v i ) executes the following steps:

[0073] v i .children→v j .children

[0074] g j =g j +g i

[0075] Remove(v i )

[0076] Firstly, the child nodes of the node v i are inserted into the child nodes of the node v j , then the g j attribute of the node v j is updated, and finally the node v i is removed from the coding tree T i , wherein v j is the parent node of v i .

[0077] S2324, Update node v j height v j .height and coding tree T i height v r The formula for updating the height property of a node is as follows:

[0078] v j .height = Max(v j .children.height)+1

[0079] v r .height = Max(v r .children.height)+1

[0080] S2325, If the coding tree T i If the height is greater than the input encoding tree height parameter k, jump to S2321.

[0081] Furthermore, in the coding tree T i In the context, for any leaf node v i If it is related to its parent node v j The difference in the height attribute is greater than the second threshold. For the second threshold, this embodiment uses 1 as an example for illustration, i.e., v j .height-v i If height > 1, then create a new node v. e , will v e Insert into node v i and node v j between.

[0082] S2331. To avoid cross-layer links, detect the coding tree T. i leaf node v i Is it related to its parent node v? j The difference in the height attribute is greater than 1; if the difference is greater than 1, continue execution downwards;

[0083] S2332, Create a new node v e ;

[0084] S2333, Insert node v e The steps are as follows:

[0085] v i →v e .children

[0086] v e →v j .children

[0087] node v i Insert into node v e In the child nodes of , v e Insert into root node v j In the child nodes;

[0088] S2334, Update v e The node's attributes are defined by the following formula:

[0089] g e =g i

[0090] v e .height=v i .height+1

[0091] S2335, If the coding tree T i leaf node v i Its parent node v j If the difference in the height attribute is greater than 1, jump to S2331.

[0092] Through step S23, the coding tree T i Transform a Huffman tree into a coding tree of height k, where the height of the root node is v. r .height equals k, the input atoms in step S1 are leaf nodes, and all leaf nodes are in the same layer, at the bottom layer.

[0093] S3. Calculate the optimal coding tree for each coding tree in the second coding tree using the information structure entropy minimization algorithm, and construct the optimal coding tree set, including:

[0094] S31. For any encoded tree, calculate the information structure entropy based on the number of edges and the relative positions between nodes, using the following formula:

[0095]

[0096] Where v t It is T i Non-root nodes in It is v t The parent node, Indicates with v t The number of edges connected to leaf nodes in a subtree that is a non-root node; vol(V) and vol(v) t ) represent V and v respectively t The sum of the in-degree weights of the middle leaf nodes.

[0097] S32, any T in the coding tree T = {T1, T2, …} generated in step S2 i Because it has a fixed number of layers, all the coding trees T i The minimum information structure entropy calculation is performed to generate the optimal coding tree, and the calculation equation is:

[0098]

[0099] After steps S2 and S3, any T in the coding tree T = {T1, T2, …} i is a coding tree with a height of k and has the minimum information structure entropy.

[0100] S4, input the optimal coding tree set into the pre-constructed graph convolutional neural network model, output the predicted label, and obtain the classification result of the protein structure according to the predicted label.

[0101] In some embodiments, the construction method of the graph convolutional neural network model comprises the following:

[0102] The optimal coding tree set is randomly divided into a training set, a validation set, and a test set, and the training set, the validation set, and the test set are normalized and preprocessed;

[0103] The graph convolutional neural network model is constructed, and the training set is used to train the graph convolutional neural network model, the validation set is used to verify the graph convolutional neural network model, and it is judged whether the graph convolutional neural network model meets the update iteration condition;

[0104] In response to meeting the update iteration condition, the test set is input into the graph convolutional neural network model for feature extraction, and the classification label prediction value corresponding to the protein structure node is output.

[0105] It should be noted that the optimal coding tree set is divided into a training set, a validation set, and a test set according to a ratio of 7:2:1 in this embodiment, a data set is constructed, and a ten-fold cross-validation method is used to evaluate the graph convolutional neural network model.

[0106] In some embodiments, the optimal coding tree set is input into the pre-constructed graph convolutional neural network model, the predicted label is output, and the classification result of the protein structure is obtained according to the predicted label, which comprises:

[0107] The graph convolutional neural network model comprises a three-layer hierarchical pooling architecture, and each layer of the hierarchical pooling architecture comprises a convolutional layer and a pooling layer.

[0108] In the convolutional layer, the input graph is convolved using a convolution formula, and the convolution formula is:

[0109]

[0110] where h l is the node representation of the l-th layer, Θ ∈ R X*Y is the convolutional weight of X-dimensional input features and Y-dimensional output features;

[0111] In the pooling layer, the input optimal coding tree set is collected, and the three-layer hierarchical pooling architecture is used for convolutional pooling on all coding trees in the optimal coding tree set to obtain the features of the coding trees. In the pooling layer, the maximum pooling method is used, which is a conventional technical means known in the art and will not be described in detail here.

[0112] The output from the pooling layer is input to the readout layer for summarization, and the summarized output is transmitted to the linear layer for classification to predict and summarize the labels of the protein structure nodes and obtain the protein structure classification result.

[0113] In some embodiments, the parameter information of the graph convolutional neural network model includes:

[0114] The number of convolutional layers, the height of the coding tree, the size of the input batch during training, the number of training rounds, the learning rate, the random seed, and the dropout rate of the last layer.

[0115] Compared with the prior art, the method uses a global optimization method and does not need to be learned. The clustering method of hierarchical pooling can be directly obtained from the minimum information structure entropy algorithm, and the method relies on compression of the number of layers, which is beneficial to retaining the local structure of the graph and can efficiently classify and detect protein structures. The model used can greatly improve the prediction speed of protein structures.

[0116] The protein structure classification and detection method based on the minimum information structure entropy provided in the embodiment generates a coding tree by processing the growth trajectory data of the acquired protein molecules, calculates the coding tree by the minimum information structure entropy algorithm to obtain an optimal coding tree set, and uses a graph convolutional neural network model to predict and classify the protein structure to obtain the classification result of the protein molecule structure. The method improves the prediction speed of complex protein molecules, improves the accuracy and efficiency of protein molecule classification, and has important significance for protein research and drug development.

[0117] It should be noted that some embodiments of the present application have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than that described above and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous or possible.

[0118] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the embodiments of the present application also provide a protein structure classification detection device based on minimizing information structure entropy.

[0119] As shown in Figure 2 , the protein structure classification detection device based on minimizing information structure entropy comprises:

[0120] The acquisition module 11 is configured to acquire growth trajectory data of a protein molecule and construct a directed graph network set;

[0121] The encoding tree generation module 12 is configured to initialize the directed graph network set as a first encoding tree, calculate information structure entropy of leaf nodes of the first encoding tree, and generate a second encoding tree with a height of k;

[0122] The optimal encoding tree construction module 13 is configured to calculate each encoding tree in the second encoding tree by an information structure entropy minimization algorithm, obtain an optimal encoding tree corresponding to each encoding tree, and construct an optimal encoding tree set;

[0123] The result classification module 14 is configured to input the optimal encoding tree set into a pre-constructed graph convolutional neural network model, output a predicted label, and obtain a classification result of the protein structure according to the predicted label.

[0124] For the convenience of description, the above device is described as various modules respectively described in terms of functions. Of course, when implementing the embodiments of the present application, the functions of the modules can be implemented in the same or multiple software and / or hardware.

[0125] The device of the above embodiments is used to implement the corresponding protein structure classification detection method based on minimizing information structure entropy in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which are not described here.

[0126] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the embodiments of the present application also provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the protein structure classification detection method based on minimizing information structure entropy as described in any of the above embodiments when executing the program.

[0127] Figure 3A more specific electronic device hardware structure schematic diagram provided by the embodiment is shown, and the device can include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 for internal communication.

[0128] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, etc., for executing related programs to implement the technical solutions provided by the embodiments of the present specification.

[0129] The memory 1020 can be implemented by a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 1020 can store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present specification are implemented by software or firmware, the related program codes are stored in the memory 1020 and called and executed by the processor 1010.

[0130] The input / output interface 1030 is used to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0131] The communication interface 1040 is used to connect a communication module (not shown in the figure) to realize the communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as USB, network cable, etc.) or through a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0132] The bus 1050 includes a channel for transmitting information between various components (such as the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040) of the device.

[0133] It should be noted that although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can also only contain components necessary to implement the embodiments of the present application, and does not necessarily contain all the components shown in the figure.

[0134] The electronic device of the above embodiment is used to implement the corresponding protein structure classification detection method based on minimizing information structural entropy in any of the preceding embodiments, and has the beneficial effects of the corresponding method embodiments, which are not repeated here.

[0135] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the present application also provides a non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the protein structure classification detection method based on minimizing information structural entropy as described in any of the above embodiments.

[0136] The computer readable medium of the present embodiment includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0137] The computer instructions stored in the storage medium of the above embodiment are used to cause the computer to perform the protein structure classification detection method based on minimizing information structural entropy as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which are not repeated here.

[0138] Those of ordinary skill in the art will realize that the foregoing discussion of any of the embodiments has been presented for the purpose of illustration and description and is not intended to be exhaustive or to limit the application to the precise forms described, and that various adaptations and modifications are possible within the scope and spirit of the application. For example, while the embodiments discussed above have been described in the context of a memory device, other memory architectures (e.g., dynamic RAM (DRAM)) can use the embodiments discussed.

[0139] In addition, to simplify the description and discussion, and so as not to make the embodiments of the application difficult to understand, the well-known power / ground connections to integrated circuit (IC) chips and other components can or can not be shown in the provided drawings. Further, devices can be shown in block diagram form so as not to make the embodiments of the application difficult to understand, and this also takes into account the fact that details regarding implementation of these block diagram devices are highly dependent on the platform in which the embodiments of the application are to be implemented (i.e., these details should be well within the understanding of one of ordinary skill in the art). Where specific details (e.g., circuitry) are set forth in order to describe an illustrative embodiment of the application, it should be apparent to one of ordinary skill in the art that the embodiments of the application can be practiced without or with variations of these specific details. Thus, the description should not be viewed as limiting the application, but rather as merely describing illustrative embodiments.

[0140] While the application has been described in connection with specific embodiments thereof, it will be understood that many modifications, variations and alternatives will be apparent to those skilled in the art as a result of the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) can use the embodiments discussed.

[0141] It is therefore intended that the embodiments of the application embrace all such alternatives, modifications and variations as falling within the broad scope of the appended claims. Accordingly, any and all departures from the above described embodiments are intended to be included within the scope of the application as defined by the following claims.

Claims

1. A method for protein structure classification detection based on minimizing the entropy of information structure, characterized in that, The method comprises: obtaining growth trajectory data of protein molecules and constructing a directed graph network set; initializing the directed graph network set as a first encoding tree, calculating information structure entropy of leaf nodes of the first encoding tree, and generating a second encoding tree with a height of k, comprising: performing node merging entropy calculation, node removal entropy calculation and node height comparison on the leaf nodes of the first encoding tree to generate a second encoding tree with a height of k, comprising: For any coding tree T i , using a node merging entropy calculation formula to select nodes v i and v j from leaf nodes of a root node v r , create a new node v e , and insert node v e between v r and (v i , v j ); after all leaf nodes are merged, if a value of a height attribute height of the root node v r is greater than a first threshold, merge the leaf nodes and their parent nodes until the value of the height attribute height of the root node v r is equal to the first threshold, wherein the node merging entropy calculation formula is: where v i is the weighted in-degree of node v i , g i is the sum of the weighted in-degree of node v i and the weighted in-degree of its child nodes, (v i -g i ) is the weighted in-degree of the child nodes of node v i , v is the sum of the weighted in-degree of node v i and the weighted in-degree of node v j , v j is the weighted in-degree of node v j , g j is the sum of the weighted in-degree of node v j and the weighted in-degree of its child nodes, (v j -g j ) is the weighted in-degree of the child nodes of node v j ; and vol(V) is the sum of the weighted in-degree of the entire tree; denotes obtaining (v i , v j ) with the minimum value; calculating each encoding tree in the second encoding tree by an information structure entropy minimization algorithm to obtain an optimal encoding tree corresponding to each encoding tree, and constructing an optimal encoding tree set; inputting the optimal encoding tree set into a pre-constructed graph convolutional neural network model to output a prediction label, and obtaining a classification result of the protein structure according to the prediction label.

2. The method of claim 1, wherein the method is based on minimizing the entropy of the information structure of the protein structure classification. The method comprises: obtaining growth trajectory data of protein molecules and constructing a directed graph network set; obtaining growth trajectory data of protein molecules and constructing a directed graph network set; According to the growth order of the protein molecules, the direction of the edge in the directed graph network is determined; 3. The method of claim 1, wherein the method is based on minimizing the entropy of the information structure of the protein structure classification. The directed graph network generated by each protein molecule is integrated into a directed graph network set. The method comprises: where v t is a non-root node in a coding tree T i , is a parent node of v t , denotes the number of edges connected to leaf nodes in a sub-tree with v t as the non-root node, and vol(V) and vol(v t ) denote the in-degree weights of leaf nodes in V and v t , respectively. The minimum information structure entropy calculation is performed on the coding tree T i to generate the corresponding optimal coding tree T i The minimum information structure entropy calculation formula is as follows:

4. The method of claim 1, wherein the method is based on minimizing the entropy of the information structure of the protein structure classification. For each encoding tree, the information structure entropy is calculated according to the number of adjacent edges of the leaf nodes of the encoding tree and the relative position between the leaf nodes, and the formula is as follows: The method for constructing a graph convolutional neural network model comprises: randomly dividing the optimal encoding tree set into a training set, a validation set and a test set, and performing normalization preprocessing on the training set, the validation set and the test set; constructing a graph convolutional neural network model, training the graph convolutional neural network model using the training set, verifying the graph convolutional neural network model using the validation set, and determining whether the graph convolutional neural network model meets an update iteration condition; 5. The method for detecting protein structure classification based on minimizing entropy of information structure according to claim 4, characterized in that, in response to meeting the update iteration condition, inputting the test set into the graph convolutional neural network model for feature extraction, and outputting a classification label prediction value corresponding to a protein structure node. The method comprises: The graph convolutional neural network model comprises a three-layer hierarchical pooling architecture, and each layer of the hierarchical pooling architecture comprises a convolution layer and a pooling layer; where h l is the node representation of the lth layer, Θ ∈ R X*Y is the convolution weight of X-dimensional input features and Y-dimensional output features; In the convolution layer, the input graph is convolved by a convolution formula, and the convolution formula is as follows: In the pooling layer, the optimal encoding tree set is input, and all encoding trees in the optimal encoding tree set are convolved and pooled using a three-layer hierarchical pooling architecture to obtain features of the encoding trees; The output from the pooling layer is input to a readout layer for summarization, and the summarized output is input to a linear layer for classification to predict and summarize the labels of the protein structure nodes, thereby obtaining a protein structure classification result.

6. The method for detecting protein structure classification based on minimizing entropy of information structure according to claim 5, characterized in that, The parameter information of the graph convolutional neural network model includes: The number of convolutional layers, the height of the encoding tree, the size of the input batch during training, the number of training rounds, the learning rate, the random seed, and the dropout rate of the last layer.

7. A protein structure classification detection apparatus based on minimizing the entropy of information structure, characterized by, The device comprises: An acquisition module configured to acquire growth trajectory data of a protein molecule and construct a directed graph network set; An encoding tree generation module configured to initialize the directed graph network set as a first encoding tree, calculate information structure entropy for leaf nodes of the first encoding tree, and generate a second encoding tree with a height of k, including: Performing node merging entropy calculation, node removal entropy calculation, and node height comparison on the leaf nodes of the first encoding tree to generate a second encoding tree with a height of k, including: For any coding tree T i , select nodes v r and v i from leaf nodes of root node v j , create a new node v e , insert node v e between v r and (v i , v j ); after all leaf nodes are merged, if the value of height attribute of root node v r is greater than a first threshold, merge leaf nodes and their parent nodes until the value of height attribute of root node v r is equal to the first threshold, wherein the formula for calculating node merge entropy is: where v i is the weighted in-degree sum of node v i , g i is the sum of the weighted in-degree sum of node v i and the weighted in-degree sum of child nodes, (v i -g i ) is the weighted in-degree sum of child nodes of node v i , v is the sum of the weighted in-degree sum of node v i and the weighted in-degree sum of node v j , v i is the weighted in-degree sum of node v j , g j is the sum of the weighted in-degree sum of node v j and the weighted in-degree sum of child nodes, (v j -g j ) is the weighted in-degree sum of child nodes of node v j ; and vol(V) is the sum of the weighted in-degree of the whole tree; denotes obtaining (v i , v j ) with the minimum value; An optimal encoding tree construction module configured to calculate each encoding tree in the second encoding tree by an information structure entropy minimization algorithm to obtain an optimal encoding tree corresponding to each encoding tree, and construct an optimal encoding tree set; A result classification module configured to input the optimal encoding tree set into a pre-constructed graph convolutional neural network model, output a predicted label, and obtain a classification result of a protein structure according to the predicted label.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the protein structure classification detection method based on information structure entropy minimization according to any one of claims 1-6 when executing the program.

9. A non-transitory computer-readable storage medium, comprising: Wherein, The non-transitory computer readable storage medium stores computer instructions for causing the computer to execute the protein structure classification detection method based on information structure entropy minimization according to any one of claims 1-6.

Images