A lncrna-protein interaction prediction method and a related device thereof

Abstract

The embodiment of the application belongs to the technical field of digital medical treatment, is applied to the scene of lncRNA and protein interaction prediction, and relates to a lncRNA-protein interaction prediction method and related equipment thereof, which comprises the following steps: constructing a first association graph according to lncRNA sequences; constructing a second association graph according to protein sequences; obtaining target matrices corresponding to the first association graph and the second association graph respectively; obtaining group feature vectors based on the target matrices corresponding to the first association graph and the second association graph respectively; and predicting whether there is interaction between lncRNA sequences and protein sequences in each sequence group according to the group feature vectors. The lncRNA sequences and the protein sequences are respectively trained by using a graph autoencoder training mode, and then a predictor is trained in combination with the group feature vectors, so that the structural information of the lncRNA sequences and the protein sequences and the interaction information between the lncRNA sequences and the protein sequences are fully utilized, the effectiveness and synergy of known information utilization are improved, and the prediction accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of digital medical technology and is applied to the prediction of lncRNA-protein interactions, particularly to a method for predicting lncRNA-protein interactions and related equipment. Background Technology

[0002] With the development of the computer industry and artificial intelligence, and the advent of the big data era, traditional medical methods are gradually transforming into digital healthcare. Long noncoding RNA (lncRNA) is a class of long noncoding RNAs longer than 200 nt, and is an important component of the noncoding genome. Numerous studies have shown that lncRNAs participate in a variety of biological processes, including DNA methylation, histone modification, post-transcriptional regulation of RNA, and protein translation regulation, and are involved in the regulation of various physiological and pathological processes.

[0003] Therefore, predicting potential lncRNA-protein interactions is crucial for disease prevention and treatment. lncRNA-protein interaction prediction provides new references for hot research areas such as tumor biology and COVID-19 research. Existing methods include those based on sequence similarity or existing tags, but both of these methods involve separate prediction processes, reducing the effectiveness and synergy of utilizing known information. Summary of the Invention

[0004] The purpose of this application is to propose a method and related equipment for predicting lncRNA-protein interactions, in order to solve the problem that existing technologies cannot reasonably and effectively utilize known information in predicting potential lncRNA-protein interactions.

[0005] To address the aforementioned technical problems, this application provides a method for predicting lncRNA-protein interactions, employing the following technical solution:

[0006] A method for predicting lncRNA-protein interactions includes the following steps:

[0007] Obtain N sequence groups for interaction prediction, where each sequence group contains a lncRNA sequence and a protein sequence, and N is a positive integer;

[0008] Based on the lncRNA sequences in the N sequence groups, a sequence group association map based on lncRNA similarity is constructed as the first association map;

[0009] Based on the protein sequences in the N sequence groups, a sequence group association graph based on protein similarity is constructed as a second association graph;

[0010] The graph autoencoder in the preset interaction prediction model is used to reconstruct the first association graph and the second association graph respectively, and the target matrices corresponding to the first association graph and the second association graph are obtained respectively. The preset interaction prediction model is trained in advance based on the sequence group of known interactions.

[0011] Based on the target matrices corresponding to the first and second association graphs, the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group are concatenated to obtain the group representation vector.

[0012] The group representation vector corresponding to each sequence group is input into the predictor of the interaction prediction model to obtain the prediction results of lncRNA sequences and protein sequences in each sequence group, wherein the prediction results indicate whether there is an interaction or not.

[0013] To address the aforementioned technical problems, this application also provides a lncRNA-protein interaction prediction device, which employs the following technical solution:

[0014] A lncRNA-protein interaction prediction device, comprising:

[0015] The sequence set acquisition module is used to acquire N sequence sets to be used for interaction prediction. Each sequence set contains a lncRNA sequence and a protein sequence, where N is a positive integer.

[0016] The first association graph construction module is used to construct a sequence group association graph based on lncRNA similarity according to the lncRNA sequences in the N sequence groups, which serves as the first association graph;

[0017] The second association graph construction module is used to construct a sequence group association graph based on protein similarity according to the protein sequences in the N sequence groups, as the second association graph;

[0018] The graph autoencoder module is used to reconstruct the first association graph and the second association graph using the graph autoencoder in the preset interaction prediction model, and obtain the target matrices corresponding to the first association graph and the second association graph respectively. The preset interaction prediction model is trained in advance based on known interaction sequence groups.

[0019] The group characterization vector acquisition module is used to obtain the group characterization vector by concatenating the embedding vector corresponding to the lncRNA sequence and the embedding vector corresponding to the protein sequence in each sequence group based on the target matrices corresponding to the first association graph and the second association graph, respectively.

[0020] The predictor prediction module is used to input the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to obtain the prediction results of lncRNA sequences and protein sequences in each sequence group, wherein the prediction results indicate whether there is an interaction or not.

[0021] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0022] A computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the lncRNA-protein interaction prediction method described above.

[0023] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0024] A computer-readable storage medium storing computer-readable instructions that, when executed by a processor, implement the steps of the lncRNA-protein interaction prediction method described above.

[0025] Compared with the prior art, the embodiments of this application have the following main advantages:

[0026] The lncRNA-protein interaction prediction method described in this application involves: acquiring N sequence groups to be predicted for interaction; constructing a first association graph based on the lncRNA sequences in the N sequence groups; constructing a second association graph based on the protein sequences in the N sequence groups; reconstructing the first and second association graphs using a graph autoencoder in a preset interaction prediction model to obtain target matrices corresponding to the first and second association graphs respectively; concatenating the embedding vectors corresponding to the lncRNA and protein sequences in each sequence group based on the target matrices corresponding to the first and second association graphs respectively to obtain a group representation vector; and inputting the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to predict whether there is an interaction between the lncRNA and protein sequences in each sequence group. This application employs a collaborative training method to train graph autoencoders for lncRNA and protein sequences separately, and then combines the group representation vectors between lncRNA and protein sequences for predictor training. This approach fully utilizes the structural information of lncRNA and protein sequences themselves, as well as the interaction information between them, during training and prediction. This improves the effectiveness and synergy of utilizing known information, thereby increasing the accuracy of lncRNA-protein interaction prediction. Attached Figure Description

[0027] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0029] Figure 2 A flowchart of an embodiment of the lncRNA-protein interaction prediction method according to this application;

[0030] Figure 3 yes Figure 2 A flowchart of a specific embodiment of step 202 shown;

[0031] Figure 4 yes Figure 2 A flowchart of a specific embodiment of step 203 shown;

[0032] Figure 5 yes Figure 3 A flowchart of a specific embodiment of step 302 shown;

[0033] Figure 6 yes Figure 3 A flowchart of a specific embodiment of step 303 shown;

[0034] Figure 7 yes Figure 4 A flowchart of a specific embodiment of step 401 shown;

[0035] Figure 8 yes Figure 4 A flowchart of a specific embodiment of step 402 shown;

[0036] Figure 9 A schematic diagram of a structure of an embodiment of the lncRNA-protein interaction prediction device according to this application;

[0037] Figure 10 A schematic diagram of the structure of an embodiment of the computer device according to this application. Detailed Implementation

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0039] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0040] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0041] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0042] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0043] Terminal devices 101, 102, and 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers, etc.

[0044] Server 105 can be a server that provides various services, such as a backend server that supports the pages displayed on terminal devices 101, 102, and 103.

[0045] It should be noted that the lncRNA-protein interaction prediction method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the lncRNA-protein interaction prediction device is generally set in the server / terminal device.

[0046] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0047] Continue to refer to Figure 2 A flowchart illustrating an embodiment of the lncRNA-protein interaction prediction method according to this application is shown. The lncRNA-protein interaction prediction method includes the following steps:

[0048] Step 201: Obtain N sequence groups to be used for interaction prediction, wherein each sequence group contains a lncRNA sequence and a protein sequence, and N is a positive integer.

[0049] Step 202: Based on the lncRNA sequences in the N sequence groups, construct a sequence group association map based on lncRNA similarity as the first association map.

[0050] Continue to refer to Figure 3 , Figure 3 yes Figure 2 A flowchart of a specific embodiment of step 202 shown includes:

[0051] Step 301: The lncRNA sequence in each sequence group is segmented to obtain M lncRNA sub-sequences, where M is a positive integer.

[0052] Step 302: Calculate the sequence similarity between the M lncRNA subsequences contained in different sequence groups sequentially using the edit distance method.

[0053] Continue to refer to Figure 5 , Figure 5 yes Figure 3 A flowchart of a specific embodiment of step 302 shown includes:

[0054] Step 501: Combine the lncRNA sequences in the different sequence groups in pairs to obtain lncRNA sequence comparison groups, wherein each lncRNA sequence comparison group contains two lncRNA sequences, namely the first lncRNA sequence and the second lncRNA sequence.

[0055] Step 502: Obtain the M lncRNA sub-sequences corresponding to the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group;

[0056] Step 503: Based on the M lncRNA sub-sequences corresponding to the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group, and the edit distance method, calculate the minimum edit distance required to convert the first lncRNA sequence into the second lncRNA sequence;

[0057] Step 504: Determine the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group based on the minimum edit distance, wherein the minimum edit distance and the similarity are negatively correlated, that is, the smaller the minimum edit distance, the greater the similarity;

[0058] Step 505: Sequentially select different lncRNA sequence comparison groups as the current lncRNA sequence comparison group, and repeat steps 502 to 504 to determine the similarity between the first lncRNA sequence and the second lncRNA sequence in different lncRNA sequence comparison groups.

[0059] In this embodiment, since lncRNA is a type of long non-coding RNA with a length greater than 200 nt, directly comparing two long lncRNA sequences would result in an excessively large comparison length. Therefore, the long lncRNA sequence to be compared is first split into M lncRNA sub-sequences, and then the edit distance algorithm is used to calculate the edit distance of the lncRNA sub-sequences to obtain the minimum edit distance. The similarity between the two long lncRNA sequences to be compared is determined based on the minimum edit distance. Through splitting and editing distance algorithms, the complexity of the comparison algorithm is reduced and the comparison efficiency is improved.

[0060] Step 303: Construct the sequence group association map based on lncRNA similarity according to the sequence similarity between the M lncRNA subsequences contained in different sequence groups.

[0061] Continue to refer to Figure 6 , Figure 6 yes Figure 3 A flowchart of a specific embodiment of step 303 shown includes:

[0062] Step 601: Pre-set N association nodes equal to the number of different sequence groups, wherein each association node represents a sequence group;

[0063] Step 602: Determine whether the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group meets the preset first similarity threshold;

[0064] Step 603: If the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group does not meet the preset first similarity threshold, then reset the current lncRNA sequence comparison group and continue to execute step 602.

[0065] Step 604: If the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group meets the preset first similarity threshold, then construct node connections for the associated nodes corresponding to the first lncRNA sequence and the second lncRNA sequence respectively, reset the current lncRNA sequence comparison group, and continue to execute step 602.

[0066] Step 605: Stop executing step 602 until all lncRNA sequence alignment groups have been set as the current lncRNA sequence alignment group, and obtain the node connections between the N associated nodes as the sequence group association graph based on lncRNA similarity.

[0067] In this embodiment, a sequence group association graph is constructed based on the similarity of lncRNA sequences. That is, a sequence group association graph is constructed by comparing lncRNA sequences in different sequence groups to obtain a first association graph. In essence, the node connection is processed between the N association nodes according to the similarity between two lncRNA sequences to obtain a node connection graph as the first association graph, which makes full use of the sequence structure of the lncRNA sequence itself.

[0068] Step 203: Based on the protein sequences in the N sequence groups, construct a sequence group association graph based on protein similarity as a second association graph.

[0069] Continue to refer to Figure 4 , Figure 4 yes Figure 2 A flowchart of a specific embodiment of step 203 shown includes:

[0070] Step 401: The local sequence comparison algorithm is used to calculate the sequence similarity between protein sequences contained in different sequence groups in turn.

[0071] Continue to refer to Figure 7 , Figure 7 yes Figure 4 A flowchart of a specific embodiment of step 401 shown includes:

[0072] Step 701: Combine the protein sequences in the different sequence groups in pairs to obtain protein sequence comparison groups, wherein each protein sequence comparison group contains two protein sequences, namely the first protein sequence and the second protein sequence.

[0073] Step 702: Based on the local sequence comparison algorithm, identify similar molecular structure regions between the first protein sequence and the second protein sequence in the current protein sequence comparison group, and obtain the identification result. The local sequence comparison algorithm is the Smith-Waterman algorithm, which is used to identify similar molecular structure regions between two protein sequences.

[0074] Step 703: The identification result is scored using a preset scoring method to obtain the scoring result. The preset scoring method is the BLOSUM62 matrix scoring method, which is used to evaluate the similarity of similar molecular structure regions between two protein sequences identified by the Smith-Waterman algorithm.

[0075] Step 704: Set the scoring result as the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group;

[0076] Step 705: Sequentially select different protein sequence comparison groups as the current protein sequence comparison group, and repeat steps 702 to 704 to determine the similarity between the first protein sequence and the second protein sequence in different protein sequence comparison groups.

[0077] In this embodiment, since the protein sequence itself is composed of different amino acid structures, the Smith-Waterman algorithm is used to compare the local structure of protein sequences between different sequence groups. Then, the BLOSUM62 matrix scoring method is used to evaluate the similarity of similar molecular structure regions between two protein sequences identified by the Smith-Waterman algorithm, so as to obtain the similarity of protein sequences in different sequence groups.

[0078] Step 402: Construct the sequence group association map based on protein similarity according to the sequence similarity between protein sequences contained in different sequence groups.

[0079] Continue to refer to Figure 8 , Figure 8 yes Figure 4A flowchart of a specific embodiment of step 402 shown includes:

[0080] Step 801: Determine whether the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group meets the preset second similarity threshold;

[0081] Step 802: If the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group does not meet the preset second similarity threshold, then reset the current protein sequence comparison group and continue to execute step 801.

[0082] Step 803: If the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group meets the preset second similarity threshold, then construct node connections for the associated nodes corresponding to the first protein sequence and the second protein sequence respectively, reset the current protein sequence comparison group, and continue to execute step 801.

[0083] Step 804: Stop executing step 801 until all protein sequence alignment groups have been set as the current protein sequence alignment group, and obtain the node connections between the N associated nodes as the sequence group association graph based on protein similarity.

[0084] In this embodiment, a sequence group association graph is constructed based on protein similarity. That is, a sequence group association graph is constructed by comparing protein sequences in different sequence groups individually, and a second association graph is obtained. In essence, the node connection processing is performed between the N association nodes according to the similarity between two protein sequences, and a node connection graph is obtained as the second association graph, which makes full use of the sequence structure of the protein sequence itself.

[0085] In this embodiment, a node connection graph is obtained by connecting the N associated nodes based on the similarity between two lncRNA sequences to form the first association graph, and a node connection graph is obtained by connecting the N associated nodes based on the similarity between two protein sequences to form the second association graph. This collaborative processing method constructs the first and second association graphs respectively, effectively and fully utilizing the data information in the sequence set to be tested, and ensuring the accuracy of the prediction results.

[0086] Step 204: Use the graph autoencoder in the preset interaction prediction model to reconstruct the first association graph and the second association graph respectively, and obtain the target matrices corresponding to the first association graph and the second association graph respectively. The preset interaction prediction model is trained in advance based on the sequence group of known interactions.

[0087] In this embodiment, the step of using a graph autoencoder in a preset interaction prediction model to reconstruct the first association graph and the second association graph respectively, and obtaining the target matrices corresponding to the first association graph and the second association graph respectively, specifically includes: constructing a first adjacency matrix based on the node connections between the N associated nodes in the first association graph; constructing a second adjacency matrix based on the node connections between the N associated nodes in the second association graph; inputting the first adjacency matrix and the second adjacency matrix as inputs into the graph autoencoder, wherein the graph autoencoder includes an encoding layer based on a GCN graph convolutional network and a decoding layer based on an inner-product algorithm; obtaining the embedding vectors corresponding to each associated node in the first adjacency matrix and the embedding vectors corresponding to each associated node in the second adjacency matrix based on the encoding layer; reconstructing the first adjacency matrix based on the decoding layer and the embedding vectors corresponding to each associated node in the first adjacency matrix to obtain the target matrix corresponding to the first association graph; and reconstructing the second adjacency matrix based on the decoding layer and the embedding vectors corresponding to each associated node in the second adjacency matrix to obtain the target matrix corresponding to the second association graph.

[0088] Specifically, based on the node connections between the N associated nodes in the first association graph, a first adjacency matrix is ​​constructed. Essentially, if a node connection exists between two associated nodes, the relationship vector between the two nodes is set to 1; if no node connection exists, the relationship vector is set to 0. Similarly, based on the node connections between the N associated nodes in the second association graph, a second adjacency matrix is ​​constructed, which is also an adjacency matrix composed of 1s and 0s. The graph autoencoder separately encodes lncRNA and protein sequences, and integrates known information for collaborative training to optimize the model. This allows the graph autoencoder to mutually promote and constrain each other when encoding lncRNA and protein sequences, improving model prediction performance.

[0089] In this embodiment, before performing the step of reconstructing the first association graph and the second association graph using the graph autoencoder in the preset interaction prediction model to obtain the target matrices corresponding to the first association graph and the second association graph respectively, the method further includes: obtaining known interaction sequence groups from a preset database, dividing the dataset into a training set and a validation set, wherein the preset database includes the NPInter database and the lncPro database, and the known interaction sequence groups represent known lncRNA sequences and protein sequences interacting with each other; obtaining the group representation vector corresponding to each sequence group in the training set and the group representation vector corresponding to each sequence group in the validation set through the graph autoencoder in the interaction prediction model; and based on each sequence group in the training set... The corresponding group representation vectors are used to initially train the predictor of the interaction prediction model to obtain the initially trained predictor. Based on the group representation vectors corresponding to each sequence group in the validation set, the initially trained predictor is iteratively optimized and trained. The interaction prediction model is considered trained when the loss value meets a preset loss condition. The preset loss condition is achieved by optimizing the parameters of the interaction prediction model using a backpropagation algorithm that minimizes the loss value, performing iterative optimization training until the number of iterations reaches a preset maximum. The loss value of the graph autoencoder and the loss value of the predictor are obtained before each iteration. The sum of these two values ​​is calculated, and by comparison, the model parameters at which the sum is minimized are used as the parameters of the trained interaction prediction model.

[0090] The NPInter database is a non-coding RNA interaction data resource platform that covers comprehensive and multi-dimensional interactions between non-coding RNAs and proteins, RNA, and the genome. It provides richer interaction and molecular functional annotations and offers new references for research in hot research areas such as tumor biology and COVID-19. The lncPro database, developed by the Peking University Health Science Center, is a data platform used to predict lncRNA-protein interactions. It covers known interactions between non-coding RNAs and proteins. The sequence sets of these known interactions can serve as action tags or identification codes corresponding to known non-coding RNA-protein interactions.

[0091] In essence, for obtaining the group representation vector corresponding to each sequence group in the training set and the group representation vector corresponding to each sequence group in the validation set through the graph autoencoder in the interaction prediction model, this embodiment still adopts the method of separately obtaining lncRNA sequences and protein sequences in different sequence groups, and constructing the first association graph and the second association graph corresponding to all sequence groups in the training set, respectively. Similarly, the first association graph and the second association graph corresponding to all sequence groups in the validation set are constructed, and the group representation vector is obtained based on the first association graph and the second association graph corresponding to the same sequence group. At this time, in the model training stage, a combination of co-training and graph autoencoder is used for model training, which makes fuller use of the data information of the lncRNA sequence and protein sequence itself.

[0092] During the model validation phase, the minimum loss backpropagation algorithm was used to optimize the parameters of the interaction prediction model. Since the interaction prediction model includes a graph autoencoder and a predictor, the influencing factor of the model's loss value is the sum of the losses of both. By pre-setting a maximum number of iterations, and then comparing the sum of the losses of both after each iteration, the parameters of the model with the minimum value are selected as the parameters of the trained interaction prediction model. Of course, to ensure the accuracy of the model, the calculation of the loss value can also be upgraded by transformation. For example, the minimum loss value can be obtained by combining the loss values ​​of the first association graph, the second association graph, and the predictor during the training and validation phase. Alternatively, the minimum loss value can be obtained by combining the loss values ​​of the encoder, the decoder, and the predictor in the graph autoencoder during the training and validation phase. In short, the dimensionality upgrade or reduction of the loss value calculation mainly depends on whether the loss value of the graph autoencoder is more finely divided.

[0093] By conducting model validation training and using the minimum loss backpropagation algorithm to determine the parameters of the interaction prediction model, the accuracy of subsequent predictions was ensured. Furthermore, the training process also employed a combination of co-training and graph autoencoder methods, which made fuller use of the data information of the lncRNA sequence and protein sequence itself.

[0094] Step 205: Based on the target matrices corresponding to the first association graph and the second association graph respectively, the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group are concatenated to obtain the group representation vector.

[0095] In this embodiment, before executing the step of concatenating the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group based on the target matrices corresponding to the first association graph and the second association graph respectively, the method further includes: obtaining the embedding vectors corresponding to each lncRNA sequence in the N sequence groups through the target matrix corresponding to the first association graph; and obtaining the embedding vectors corresponding to each protein sequence in the N sequence groups through the adjacency matrix corresponding to the second association graph.

[0096] In this embodiment, the step of concatenating the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group based on the target matrices corresponding to the first and second association graphs to obtain the group representation vector specifically includes: obtaining and concatenating the embedding vectors corresponding to the lncRNA sequences and the protein sequences in the current sequence group as the group representation vector of the current sequence group; and sequentially using different sequence groups as the current sequence group to obtain the group representation vectors corresponding to different sequence groups.

[0097] Step 206: Input the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to obtain the prediction results of lncRNA sequences and protein sequences in each sequence group, wherein the prediction results indicate whether there is an interaction or not.

[0098] After performing the step of inputting the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to predict whether there is an interaction between the lncRNA sequence and the protein sequence in each sequence group, the method includes: acquiring and parsing the prediction results output by the predictor of the interaction prediction model; if the parsing shows that there is an interaction between the lncRNA sequence and the protein sequence in the current sequence group, then the current sequence group is marked for distinction and the interaction field information is output; otherwise, a prompt message indicating unknown interaction is sent to the target monitoring terminal.

[0099] This application obtains N sequence groups for interaction prediction; constructs a first association graph based on the lncRNA sequences in the N sequence groups; constructs a second association graph based on the protein sequences in the N sequence groups; reconstructs the first and second association graphs using a pre-defined graph autoencoder in the interaction prediction model, obtaining target matrices corresponding to the first and second association graphs respectively; based on the target matrices corresponding to the first and second association graphs respectively, concatenates the embedding vectors corresponding to the lncRNA and protein sequences in each sequence group to obtain a group representation vector; inputs the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to predict whether there is an interaction between the lncRNA and protein sequences in each sequence group. This application uses a co-training method to train the graph autoencoder for the lncRNA and protein sequences respectively, and then combines the group representation vector between the lncRNA and protein sequences for predictor training. This makes full use of the structural information of the lncRNA and protein sequences themselves and the interaction information between them during training and prediction, improving the effectiveness and synergy of using known information, thereby improving the accuracy of lncRNA-protein interaction prediction.

[0100] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0101] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0102] In this embodiment, the lncRNA sequence and protein sequence are trained using a collaborative training method to train a graph autoencoder, and then the predictor is trained by combining the group representation vector between the lncRNA sequence and protein sequence. This makes full use of the structural information of the lncRNA sequence and protein sequence itself and the interaction information between them during training and prediction, thereby improving the effectiveness and synergy of the use of known information.

[0103] Further reference Figure 9 As a response to the above Figure 2To implement the method shown, this application provides an embodiment of a lncRNA-protein interaction prediction device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0104] like Figure 9 As shown, the lncRNA-protein interaction prediction device 900 described in this embodiment includes: a sequence set acquisition module 901, a first association graph construction module 902, a second association graph construction module 903, a graph autoencoder module 904, a set representation vector acquisition module 905, and a predictor prediction module 906. Wherein:

[0105] The sequence set acquisition module 901 is used to acquire N sequence sets to be used for interaction prediction, wherein each sequence set contains a lncRNA sequence and a protein sequence, and N is a positive integer;

[0106] The first association graph construction module 902 is used to construct a sequence group association graph based on lncRNA similarity according to the lncRNA sequences in the N sequence groups, as the first association graph;

[0107] The second association graph construction module 903 is used to construct a sequence group association graph based on protein similarity according to the protein sequences in the N sequence groups, as the second association graph;

[0108] The graph autoencoder encoding module 904 is used to reconstruct the first association graph and the second association graph using the graph autoencoder in the preset interaction prediction model, and obtain the target matrices corresponding to the first association graph and the second association graph respectively. The preset interaction prediction model is trained in advance based on known interaction sequence groups.

[0109] The group characterization vector acquisition module 905 is used to obtain the group characterization vector by splicing the embedding vector corresponding to the lncRNA sequence and the embedding vector corresponding to the protein sequence in each sequence group based on the target matrices corresponding to the first association graph and the second association graph, respectively.

[0110] The predictor prediction module 906 is used to input the group characterization vector corresponding to each sequence group into the predictor of the interaction prediction model to obtain the prediction results of lncRNA sequences and protein sequences in each sequence group, wherein the prediction results indicate whether there is an interaction or not.

[0111] In some specific embodiments of this application, the lncRNA-protein interaction prediction device 900 further includes an interaction prediction model training module. This module is used to: acquire known interaction sequence groups from a preset database and divide them into a training set and a validation set; acquire group representation vectors corresponding to each sequence group in the training set and the validation set using a graph autoencoder in the interaction prediction model; perform initial training on the predictor of the interaction prediction model based on the group representation vectors corresponding to each sequence group in the training set to obtain a pre-trained predictor; and perform training based on the... The group representation vector corresponding to each sequence group in the validation set is used to iteratively optimize the initially trained predictor; the interaction prediction model is trained until the loss value of the interaction prediction model meets a preset loss condition, wherein the preset loss condition is to optimize the parameters of the interaction prediction model by minimizing the loss value through backpropagation algorithm, and to perform iterative optimization training until the number of iterations reaches a preset maximum number. The loss value of the graph autoencoder and the loss value of the predictor are obtained before each iteration of training, and the sum of the two is calculated. By comparison, the model parameters when the sum of the two is the minimum value are obtained as the parameters of the trained interaction prediction model.

[0112] This application obtains N sequence groups for interaction prediction; constructs a first association graph based on the lncRNA sequences in the N sequence groups; constructs a second association graph based on the protein sequences in the N sequence groups; reconstructs the first and second association graphs using a pre-defined graph autoencoder in the interaction prediction model, obtaining target matrices corresponding to the first and second association graphs respectively; based on the target matrices corresponding to the first and second association graphs respectively, concatenates the embedding vectors corresponding to the lncRNA and protein sequences in each sequence group to obtain a group representation vector; inputs the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to predict whether there is an interaction between the lncRNA and protein sequences in each sequence group. This application uses a co-training method to train the graph autoencoder for the lncRNA and protein sequences respectively, and then combines the group representation vector between the lncRNA and protein sequences for predictor training. This makes full use of the structural information of the lncRNA and protein sequences themselves and the interaction information between them during training and prediction, improving the effectiveness and synergy of using known information, thereby improving the accuracy of lncRNA-protein interaction prediction.

[0113] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0114] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0115] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 10 , Figure 10 This is a basic structural block diagram of the computer device in this embodiment.

[0116] The computer device 10 includes a memory 10a, a processor 10b, and a network interface 10c that are interconnected via a system bus. It should be noted that only the computer device 10 with components 10a-10c is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0117] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0118] The memory 10a includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10a may be an internal storage unit of the computer device 10, such as the hard disk or memory of the computer device 10. In other embodiments, the memory 10a may also be an external storage device of the computer device 10, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 10. Of course, the memory 10a may also include both the internal storage unit and the external storage device of the computer device 10. In this embodiment, the memory 10a is typically used to store the operating system and various application software installed on the computer device 10, such as computer-readable instructions for a lncRNA-protein interaction prediction method. Furthermore, the memory 10a can also be used to temporarily store various types of data that have already been output or will be output.

[0119] In some embodiments, the processor 10b may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 10b is typically used to control the overall operation of the computer device 10. In this embodiment, the processor 10b is used to execute computer-readable instructions stored in the memory 10a or to process data, for example, to execute computer-readable instructions for the lncRNA-protein interaction prediction method.

[0120] The network interface 10c may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 10 and other electronic devices.

[0121] The computer device proposed in this embodiment belongs to the field of digital medical technology and is applied to the scenario of predicting interactions between lncRNA and protein. This application obtains N sequence groups to be predicted for interaction; constructs a first association graph based on the lncRNA sequences in the N sequence groups; constructs a second association graph based on the protein sequences in the N sequence groups; uses a graph autoencoder in a preset interaction prediction model to reconstruct the first and second association graphs respectively, obtaining target matrices corresponding to the first and second association graphs respectively; based on the target matrices corresponding to the first and second association graphs respectively, concatenates the embedding vectors corresponding to the lncRNA and protein sequences in each sequence group to obtain a group representation vector; inputs the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to predict whether there is an interaction between the lncRNA and protein sequences in each sequence group. This application employs a collaborative training method to train graph autoencoders for lncRNA and protein sequences separately, and then combines the group representation vectors between lncRNA and protein sequences for predictor training. This approach fully utilizes the structural information of lncRNA and protein sequences themselves, as well as the interaction information between them, during training and prediction. This improves the effectiveness and synergy of utilizing known information, thereby increasing the accuracy of lncRNA-protein interaction prediction.

[0122] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by a processor to cause the processor to perform the steps of the lncRNA-protein interaction prediction method described above.

[0123] The computer-readable storage medium proposed in this embodiment belongs to the field of digital medical technology and is applied to the scenario of predicting interactions between lncRNA and protein. This application obtains N sequence groups to be predicted for interaction; constructs a first association graph based on the lncRNA sequences in the N sequence groups; constructs a second association graph based on the protein sequences in the N sequence groups; reconstructs the first and second association graphs using a graph autoencoder in a preset interaction prediction model to obtain target matrices corresponding to the first and second association graphs respectively; based on the target matrices corresponding to the first and second association graphs respectively, concatenates the embedding vectors corresponding to the lncRNA and protein sequences in each sequence group to obtain a group representation vector; inputs the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to predict whether there is an interaction between the lncRNA and protein sequences in each sequence group. This application employs a collaborative training method to train graph autoencoders for lncRNA and protein sequences separately, and then combines the group representation vectors between lncRNA and protein sequences for predictor training. This approach fully utilizes the structural information of lncRNA and protein sequences themselves, as well as the interaction information between them, during training and prediction. This improves the effectiveness and synergy of utilizing known information, thereby increasing the accuracy of lncRNA-protein interaction prediction.

[0124] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0125] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A method for predicting lncRNA-protein interactions, characterized in that, Includes the following steps: Obtain N sequence groups for interaction prediction, where each sequence group contains a lncRNA sequence and a protein sequence, and N is a positive integer; Based on the lncRNA sequences in the N sequence groups, a sequence group association graph based on lncRNA similarity is constructed as the first association graph. Specifically, the long lncRNA sequence in each sequence group is first split into M lncRNA sub-sequences. Then, the edit distance algorithm is used to calculate the edit distance of the lncRNA sub-sequences to obtain the minimum edit distance. The similarity between pairs of long lncRNA sequences in different sequence groups is determined based on the minimum edit distance. Based on the similarity between pairs of long lncRNA sequences, the N association nodes are connected to obtain a node connection graph as the first association graph. Based on the protein sequences in the N sequence groups, a sequence group association graph based on protein similarity is constructed as the second association graph. Specifically, the Smith-Waterman algorithm is used to compare the local structures of protein sequences in different sequence groups, and then the BLOSUM62 matrix scoring method is used to evaluate the similarity of similar molecular structure regions between two protein sequences identified by the Smith-Waterman algorithm to obtain the similarity of protein sequences in different sequence groups. Based on the similarity between two protein sequences, node connection processing is performed on the N associated nodes to obtain a node connection graph as the second association graph. The graph autoencoder in the preset interaction prediction model is used to reconstruct the first association graph and the second association graph respectively, and the target matrices corresponding to the first association graph and the second association graph are obtained respectively. The preset interaction prediction model is trained in advance based on the sequence group of known interactions. Based on the target matrices corresponding to the first and second association graphs, the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group are concatenated to obtain the group representation vector. The group representation vector corresponding to each sequence group is input into the predictor of the interaction prediction model to obtain the prediction results of lncRNA sequences and protein sequences in each sequence group, wherein the prediction results indicate whether there is an interaction or not.

2. The lncRNA-protein interaction prediction method according to claim 1, characterized in that, The step of constructing a sequence group association map based on lncRNA similarity according to the lncRNA sequences in the N sequence groups specifically includes: The lncRNA sequences in each sequence group are segmented to obtain M lncRNA subsequences, where M is a positive integer. The sequence similarity among the M lncRNA subsequences contained in different sequence groups was calculated sequentially using the edit distance method. Based on the sequence similarity among the M lncRNA subsequences contained in different sequence groups, the sequence group association map based on lncRNA similarity is constructed. The step of constructing a sequence group association map based on protein similarity from the N sequence groups specifically includes: The sequence similarity between protein sequences in different sequence groups was calculated sequentially using a local sequence comparison algorithm. Based on the sequence similarity between protein sequences contained in different sequence groups, the sequence group association map based on protein similarity is constructed.

3. The lncRNA-protein interaction prediction method according to claim 2, characterized in that, The step of calculating the sequence similarity among the M lncRNA subsequences contained in different sequence groups sequentially using the edit distance method specifically includes: Step 501: Combine the lncRNA sequences in the different sequence groups in pairs to obtain lncRNA sequence comparison groups, wherein each lncRNA sequence comparison group contains two lncRNA sequences, namely the first lncRNA sequence and the second lncRNA sequence. Step 502: Obtain the M lncRNA sub-sequences corresponding to the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group; Step 503: Based on the M lncRNA sub-sequences corresponding to the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group, and the edit distance method, calculate the minimum edit distance required to convert the first lncRNA sequence into the second lncRNA sequence; Step 504: Determine the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group based on the minimum edit distance, wherein the minimum edit distance and the similarity are negatively correlated, that is, the smaller the minimum edit distance, the greater the similarity; Step 505: Sequentially select different lncRNA sequence comparison groups as the current lncRNA sequence comparison group, and repeat steps 502 to 504 to determine the similarity between the first lncRNA sequence and the second lncRNA sequence in different lncRNA sequence comparison groups; The step of constructing the sequence set association map based on lncRNA similarity according to the sequence similarity among the M lncRNA sub-sequences contained in different sequence sets specifically includes: Step 601: Pre-set N association nodes equal to the number of different sequence groups, wherein each association node represents a sequence group; Step 602: Determine whether the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group meets the preset first similarity threshold; Step 603: If the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group does not meet the preset first similarity threshold, then reset the current lncRNA sequence comparison group and continue to execute step 602. Step 604: If the similarity between the first lncRNA sequence and the second lncRNA sequence in the current lncRNA sequence comparison group meets the preset first similarity threshold, then construct node connections for the associated nodes corresponding to the first lncRNA sequence and the second lncRNA sequence respectively, reset the current lncRNA sequence comparison group, and continue to execute step 602. Step 605: Stop executing step 602 until all lncRNA sequence alignment groups have been set as the current lncRNA sequence alignment group, and obtain the node connections between the N associated nodes as the sequence group association graph based on lncRNA similarity.

4. The lncRNA-protein interaction prediction method according to claim 3, characterized in that, The step of sequentially calculating the sequence similarity between protein sequences in different sequence groups using a local sequence comparison algorithm specifically includes: Step 701: Combine the protein sequences in the different sequence groups in pairs to obtain protein sequence comparison groups, wherein each protein sequence comparison group contains two protein sequences, namely the first protein sequence and the second protein sequence. Step 702: Based on the local sequence comparison algorithm, identify similar molecular structure regions between the first protein sequence and the second protein sequence in the current protein sequence comparison group, and obtain the identification result. The local sequence comparison algorithm is the Smith-Waterman algorithm, which is used to identify similar molecular structure regions between two protein sequences. Step 703: The identification result is scored using a preset scoring method to obtain the scoring result. The preset scoring method is the BLOSUM62 matrix scoring method, which is used to evaluate the similarity of similar molecular structure regions between two protein sequences identified by the Smith-Waterman algorithm. Step 704: Set the scoring result as the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group; Step 705: Sequentially select different protein sequence comparison groups as the current protein sequence comparison group, and repeat steps 702 to 704 to determine the similarity between the first protein sequence and the second protein sequence in different protein sequence comparison groups. The step of constructing the sequence group association map based on protein similarity according to the sequence similarity between protein sequences contained in different sequence groups specifically includes: Step 801: Determine whether the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group meets the preset second similarity threshold; Step 802: If the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group does not meet the preset second similarity threshold, then reset the current protein sequence comparison group and continue to execute step 801. Step 803: If the similarity between the first protein sequence and the second protein sequence in the current protein sequence comparison group meets the preset second similarity threshold, then construct node connections for the associated nodes corresponding to the first protein sequence and the second protein sequence respectively, reset the current protein sequence comparison group, and continue to execute step 801. Step 804: Stop executing step 801 until all protein sequence alignment groups have been set as the current protein sequence alignment group, and obtain the node connections between the N associated nodes as the sequence group association graph based on protein similarity.

5. The lncRNA-protein interaction prediction method according to claim 4, characterized in that, The step of using a graph autoencoder in a preset interaction prediction model to reconstruct the first association graph and the second association graph respectively, and obtaining the target matrices corresponding to the first association graph and the second association graph respectively, specifically includes: Based on the node connections between the N associated nodes in the first association graph, construct a first adjacency matrix; Construct a second adjacency matrix based on the node connections between the N associated nodes in the second association graph; The first adjacency matrix and the second adjacency matrix are used as inputs to the graph autoencoder, wherein the graph autoencoder includes an encoding layer based on a GCN graph convolutional network and a decoding layer based on an inner-product algorithm; The embedding vectors corresponding to each associated node in the first adjacency matrix and the embedding vectors corresponding to each associated node in the second adjacency matrix are obtained according to the coding layer. Based on the embedding vectors corresponding to each associated node of the decoding layer and the first adjacency matrix, the first adjacency matrix is ​​reconstructed to obtain the target matrix corresponding to the first association graph; Based on the embedding vectors corresponding to each associated node of the decoding layer and the second adjacency matrix, the second adjacency matrix is ​​reconstructed to obtain the target matrix corresponding to the second association graph.

6. The lncRNA-protein interaction prediction method according to claim 1 or 5, characterized in that, Before performing the step of reconstructing the first association graph and the second association graph using the graph autoencoder in the preset interaction prediction model to obtain the target matrices corresponding to the first association graph and the second association graph respectively, the method further includes: The sequence sets of known interactions are obtained from a pre-defined database, and a training set and a validation set are divided. The pre-defined database includes the NPInter database and the lncPro database. The sequence sets of known interactions represent the sequence sets of known lncRNA sequences and protein sequences that interact with each other. The graph autoencoder in the interaction prediction model is used to obtain the group representation vector corresponding to each sequence group in the training set and the group representation vector corresponding to each sequence group in the validation set. The predictor of the interaction prediction model is initially trained based on the group representation vector corresponding to each sequence group in the training set to obtain the predictor after initial training. The predictor that has been initially trained is iteratively optimized based on the group representation vector corresponding to each sequence group in the validation set. The interaction prediction model is considered complete when its loss value meets a preset loss condition. The preset loss condition is achieved by optimizing the parameters of the interaction prediction model through a backpropagation algorithm that minimizes the loss value, performing iterative optimization training until the number of iterations reaches a preset maximum. The loss values ​​of the graph autoencoder and the predictor are obtained before each iteration, and their sum is calculated. By comparison, the model parameters at which the sum is minimized are obtained as the parameters of the completed interaction prediction model.

7. The lncRNA-protein interaction prediction method according to claim 5, characterized in that, Before executing the step of concatenating the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group based on the target matrices corresponding to the first and second association graphs respectively, to obtain the group characterization vector, the method further includes: Using the target matrix corresponding to the first association graph, obtain the embedding vector corresponding to each lncRNA sequence in the N sequence groups; By using the adjacency matrix corresponding to the second association graph, the embedding vector corresponding to each protein sequence in the N sequence groups is obtained; The step of concatenating the embedding vectors corresponding to the lncRNA sequences and the protein sequences in each sequence group based on the target matrices corresponding to the first and second association graphs to obtain the group representation vector specifically includes: Obtain and concatenate the embedding vectors corresponding to the lncRNA sequences and the protein sequences in the current sequence group, and use them as the group representation vector of the current sequence group; By sequentially taking different sequence groups as the current sequence group, the group representation vectors corresponding to different sequence groups are obtained.

8. A lncRNA-protein interaction prediction device, characterized in that, include: The sequence set acquisition module is used to acquire N sequence sets to be used for interaction prediction. Each sequence set contains a lncRNA sequence and a protein sequence, where N is a positive integer. The first association graph construction module is used to construct an association graph based on lncRNA similarity from the N sequence groups, which serves as the first association graph. Specifically, the long lncRNA sequence in each sequence group is first split into M lncRNA sub-sequences. Then, the edit distance algorithm is used to calculate the edit distance of the lncRNA sub-sequences to obtain the minimum edit distance. The similarity between pairs of long lncRNA sequences in different sequence groups is determined based on the minimum edit distance. Based on the similarity between pairs of long lncRNA sequences, the N association nodes are connected to obtain a node connection graph as the first association graph. The second association graph construction module is used to construct a sequence group association graph based on protein similarity based on the protein sequences in the N sequence groups, which serves as the second association graph. Specifically, the Smith-Waterman algorithm is used to compare the local structures of protein sequences in different sequence groups, and then the BLOSUM62 matrix scoring method is used to evaluate the similarity of similar molecular structure regions between two protein sequences identified by the Smith-Waterman algorithm to obtain the similarity of protein sequences in different sequence groups. Based on the similarity between two protein sequences, node connection processing is performed on the N association nodes to obtain a node connection graph as the second association graph. The graph autoencoder module is used to reconstruct the first association graph and the second association graph using the graph autoencoder in the preset interaction prediction model, and obtain the target matrices corresponding to the first association graph and the second association graph respectively. The preset interaction prediction model is trained in advance based on known interaction sequence groups. The group characterization vector acquisition module is used to obtain the group characterization vector by concatenating the embedding vector corresponding to the lncRNA sequence and the embedding vector corresponding to the protein sequence in each sequence group based on the target matrices corresponding to the first association graph and the second association graph, respectively. The predictor prediction module is used to input the group representation vector corresponding to each sequence group into the predictor of the interaction prediction model to obtain the prediction results of lncRNA sequences and protein sequences in each sequence group, wherein the prediction results indicate whether there is an interaction or not.

9. A computer device comprising a memory and a processor, the memory storing computer-readable instructions, wherein the processor, when executing the computer-readable instructions, implements the steps of the lncRNA-protein interaction prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions that, when executed by a processor, implement the steps of the lncRNA-protein interaction prediction method as described in any one of claims 1 to 7.

Images