Method, device, terminal equipment and medium for predicting association between mRNA and drugs

Abstract

The present application is applicable to the field of bioinformatics, and provides a method, apparatus, terminal device and medium for predicting the association between mRNA and drugs. Among them, the method aggregates neighbor nodes of the vector representations of mRNA and drugs on a pre-constructed association bipartite graph to obtain the initial feature vectors of the two respectively; constructs an association hypergraph based on the association bipartite graph to obtain the hypergraph feature vector representation of mRNA and drugs; constructs a cross-view comparison loss based on the initial feature vector and the hypergraph feature vector representation to obtain a comprehensive loss; uses the comprehensive loss to update the initial feature vectors of mRNA and drugs respectively to obtain the intermediate feature vectors of mRNA and drugs; if the intermediate feature vector meets the update termination condition, uses it to construct a correlation prediction model to predict the association between mRNA and drugs. The present application can improve the accuracy of the prediction of the association between mRNA and drugs.

Description

Technical Field

[0001] The present application belongs to the field of bioinformatics technology, and in particular relates to a method, apparatus, terminal device, and medium for predicting the association between mRNA and drugs. Background Art

[0002] In recent years, due to individual heterogeneity, drug sensitivity has varied significantly between individuals, resulting in inefficient utilization and significant waste of medical resources. Furthermore, differences in drug sensitivity can necessitate different medications for different patients to achieve therapeutic efficacy. Studies have shown that alterations in biomolecules, including protein-coding messenger RNA (mRNA) expression, copy number variation, mutation, and methylation, significantly influence sensitivity to clinical drug therapy. Studies have also revealed a causal relationship between point mutations in KRAS (a murine sarcoma viral oncogene) and acquired resistance to cetuximab in colorectal cancer. Therefore, studying the mRNA-drug sensitivity association is crucial for optimizing drug therapy, achieving personalized disease treatment, and advancing precision medicine. However, biological experiments to validate the mRNA-drug sensitivity association are highly unreliable and require significant time and effort. With the development of deep learning technology, bioinformatics has attracted the attention of more and more researchers. More and more computational methods have been proposed to predict lncRNA-disease associations, microRNA (miRNA)-disease associations, circular RNA (circRNA)-disease associations, etc. However, the current prediction methods for mRNA and drug associations are rare and not very accurate. Summary of the Invention

[0003] The embodiments of the present application provide a method, apparatus, terminal device, and medium for predicting the association between mRNA and drugs, which can improve the accuracy of the prediction of the association between mRNA and drugs.

[0004] In a first aspect, the present invention provides a method for predicting the association between mRNA and drugs, comprising:

[0005] Step 1: Run an information aggregation neural network model on a pre-constructed bipartite graph of association between the test mRNA and the target drug, and perform neighbor node aggregation on the pre-initialized vector representation of the test mRNA and the pre-initialized vector representation of the target drug to obtain the initial feature vector of the test mRNA and the initial feature vector of the target drug;

[0006] Step 2: construct an association hypergraph based on the association bipartite graph to obtain the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug;

[0007] Step 3: For either the test mRNA or the target drug, a cross-view contrast loss is constructed based on the initial feature vector and hypergraph feature vector representation of the test mRNA and the target drug, respectively. The cross-view contrast loss of the test mRNA and the target drug are obtained.

[0008] Step 4: construct a comprehensive loss based on the cross-view comparison loss of the mRNA to be tested and the cross-view comparison loss of the target drug;

[0009] Step 5: Using the comprehensive loss, backpropagate and update the initial feature vectors of the mRNA to be tested and the target drug to obtain the intermediate feature vectors of the mRNA to be tested and the target drug;

[0010] Step 6: If the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet the preset update termination criteria, the intermediate feature vector of the mRNA to be tested is used as the final feature vector of the mRNA to be tested, and the intermediate feature vector of the target drug is used as the final feature vector of the target drug; otherwise,

[0011] The intermediate feature vector of the mRNA to be tested is used as the vector representation of the mRNA to be tested in step 1, and the intermediate feature vector of the target drug is used as the vector representation of the target drug in step 1, and the process returns to step 1;

[0012] Step 7: constructing a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug;

[0013] Step 8: Use the correlation prediction model to predict the correlation between the mRNA to be tested and the target drug.

[0014] Optionally, the information aggregation neural network model in step 1 is a graph convolutional neural network model.

[0015] Optionally, in step 1, neighbor node aggregation is performed on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug, including:

[0016] By calculating the formula

[0017]

[0018]

[0019]

[0020] Get the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug in, represents the initial feature vector of the i-th mRNA to be tested output by the l-th layer neural network, represents the initial feature vector of the jth target drug output by the lth layer neural network, i=1,2,...,I, I represents the total number of mRNAs pre-acquired in step 1, j=1,2,...,J, J represents the total number of drugs pre-acquired in step 1, Represents the aggregated information of neighbor nodes of the mRNA to be tested in the associated bipartite graph, Represents the neighbor node aggregation information of the target drug in the associated bipartite graph, w represents the vector dimension, R represents the real number domain, σ(·) represents the nonlinear activation function, represents the normalized adjacency matrix, D (m) Represents the angle matrix of the mRNA to be tested, D (m) ∈R I×I , d (d) Denotes the diagonal matrix of the target drug, D (d) ∈R J×J , N i Represents the central node m of the mRNA to be tested in the associated bipartite graph i Neighboring nodes, N j Represents the central node d of the target drug in the associated bipartite graph j The neighboring nodes of Represents the vector representation of the i-th mRNA to be tested output by the l-1 layer of the graph convolutional neural network model, represents the vector representation of the j-th target drug output by the l-1 layer of the graph convolutional neural network model, represents the vector representation after the aggregation of the i-th mRNA information in the l-th layer neural network, A represents the vector representation of the j-th drug information aggregation in the l-th layer neural network, i,j Represents the adjacency matrix between the i-th mRNA to be tested and the j-th target drug.

[0021] Optionally, step 2 includes:

[0022] By calculating the formula

[0023]

[0024]

[0025]

[0026] ψ(X)=σ(VX)+X

[0027] Get the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug in, It represents the hyperedge feature vector representation of the mRNA to be tested at the lth hypergraph information aggregation layer, represents the hyperedge feature vector representation of the target drug at the lth hypergraph information aggregation layer, W (m) Represents the feature matrix of the hyperedge of the mRNA to be tested, W (m) ∈R d×H , W (d) The characteristic matrix of the hyperedge representing the target drug, W (d) ∈R d×H , Λ (m) Represents the hypergraph feature vector of the mRNA to be tested, Λ (d) Represents the hypergraph feature vector of the target drug, Λ (m) The vector representing the mRNA to be tested is E (m) The hypergraph structure matrix H corresponding to the mRNA to be tested (m) Polymerization, Λ (d) The vector representation E of the target drug (d) The hypergraph structure matrix H corresponding to the target drug (d) Aggregation is obtained, Λ∈R H×d , E∈R K×d , H∈R K×H , K=I or K=J, c represents the number of hypergraph vector representation layers, ψ(·) represents the encoding function of the hypergraph information aggregation layer, X represents the parameter matrix, X∈R H×H , σ' represents the LeakyReLU mapping.

[0028] Optionally, step 3 includes:

[0029] By calculating the formula

[0030]

[0031] Get the view comparison loss of the mRNA to be tested and cross-view comparison loss of target drugs i- represents the negative sample of the i-th mRNA, and j- represents the negative sample of the j-th drug.

[0032] Optionally, step 4 includes:

[0033] By calculating the formula

[0034]

[0035]

[0036] Get the comprehensive loss L; where, represents the association score of the positive sample, represents the association score of negative samples, Θ represents the model parameters, r i represents the i-th mRNA, m j represents the jth drug.

[0037] Optionally, step 7 includes:

[0038] By calculating the formula

[0039]

[0040] Get the final vector representation of the mRNA to be tested and the final vector representation of the target drug in, represents the vector representation of the i-th mRNA information after aggregation in the l-th layer hypergraph convolutional network, Represents the vector representation of the j-th drug information after aggregation in the l-th layer hypergraph convolutional network;

[0041] according to and Construct a correlation prediction model.

[0042] Optionally, the expression of the correlation prediction model is as follows:

[0043]

[0044] in, represents the feature vector of the i-th mRNA in the l-th layer of the graph convolutional network, represents the feature vector of the jth drug in the lth layer of the graph convolutional network, represents the association score between the i-th mRNA and the j-th drug.

[0045] In a second aspect, the present invention provides a device for predicting the association between mRNA and drugs, comprising:

[0046] An initialization module is used to run an information aggregation neural network model on a pre-constructed bipartite graph of association between the mRNA to be tested and the target drug, and to perform neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug, respectively, to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug;

[0047] The hypergraph feature vector module is used to construct an association hypergraph based on the association bipartite graph to obtain the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug;

[0048] The contrast loss module is used to construct a cross-view contrast loss for either the test mRNA or the target drug based on its initial feature vector and hypergraph feature vector representation, thereby obtaining the view contrast loss of the test mRNA and the cross-view contrast loss of the target drug.

[0049] The feature vector optimization module is used to construct a comprehensive loss based on the cross-view comparison loss function of the mRNA to be tested and the cross-view comparison loss function of the target drug;

[0050] The intermediate feature vector is used to back-propagate and update the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug using the comprehensive loss, thereby obtaining the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug;

[0051] The final feature vector module is used to use the intermediate feature vector of the mRNA to be tested as the final feature vector of the mRNA to be tested and the intermediate feature vector of the target drug as the final feature vector of the target drug if the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet the preset update termination criteria; otherwise,

[0052] The intermediate feature vector of the mRNA to be tested is used as the vector representation of the mRNA to be tested in the initialization module, and the intermediate feature vector of the target drug is used as the vector representation of the target drug in the initialization module, and the initialization module is returned to execute;

[0053] A correlation prediction model module is used to construct a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug;

[0054] The prediction module is used to predict the correlation between the mRNA to be tested and the target drug using the correlation prediction model.

[0055] In a third aspect, an embodiment of the present application provides a terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned method for predicting the association between mRNA and drugs is implemented.

[0056] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned method for predicting the association between mRNA and drugs.

[0057] The above solution of the present application has the following beneficial effects:

[0058] In some embodiments of the present application, a cross-view contrast loss is constructed based on the initial feature vectors and hypergraph feature vector representations of the mRNA to be tested and the target drug, and then the initial feature vectors of the mRNA to be tested and the target drug are back-propagated and updated respectively according to the cross-view contrast loss, so that a more accurate feature vector can be obtained, thereby improving the accuracy of the prediction of the association between the mRNA to be tested and the target drug.

[0059] Other beneficial effects of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0061] Figure 1 A flowchart of a method for predicting the association between mRNA and drugs provided in one embodiment of the present application;

[0062] Figure 2 An ROC curve diagram showing the performance comparison between the mRNA and drug association prediction method provided in one embodiment of the present application and other existing technologies;

[0063] Figure 3 A PR curve diagram showing the performance comparison between the mRNA and drug association prediction method provided in one embodiment of the present application and other existing technologies;

[0064] Figure 4 A schematic diagram of the structure of a device for predicting the association between mRNA and drugs provided in one embodiment of the present application;

[0065] Figure 5 A schematic diagram of the structure of a terminal device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0066] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0067] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0068] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0069] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0070] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0071] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0072] In response to the problem that the current prediction methods for the association between mRNA and drugs are not very accurate, the present application provides a method, apparatus, terminal device and medium for predicting the association between mRNA and drugs. The method for predicting the association between mRNA and drugs constructs a cross-view comparison loss based on the initial feature vectors and hypergraph feature vector representations of the mRNA to be tested and the target drug, and then backpropagates and updates the initial feature vectors of the mRNA to be tested and the target drug based on the cross-view comparison loss, thereby obtaining a more accurate feature vector, thereby improving the accuracy of the prediction of the association between the mRNA to be tested and the target drug.

[0073] like Figure 1 As shown, the method for predicting the association between mRNA and drugs provided in this application mainly includes the following steps:

[0074] Step 1: Run the information aggregation neural network model on the pre-constructed bipartite graph of the association between the mRNA to be tested and the target drug, and perform neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug, respectively, to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug.

[0075] In an embodiment of the present application, the above-mentioned information aggregation neural network model is a graph convolutional neural network model.

[0076] It should be noted that the construction of the above-mentioned association bipartite graph is based on the association pairs consisting of the mRNA to be tested and the target drug, and can be achieved using a common bipartite graph construction method. In some embodiments of the present application, the above-mentioned association pairs can be obtained from the RNAacrDrug database (containing RNA related to drug sensitivity from multi-omics data).

[0077] To increase the credibility of the correlation prediction model, in some embodiments of the present application, the 4,198,837 association pairs obtained were divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The training set was then used to train the correlation prediction model to obtain a trained correlation prediction model. The validation set was then used to optimize the parameters of the trained correlation prediction model. Finally, the performance of the prediction model was evaluated using the test set.

[0078] The mRNA to be tested is one of multiple mRNAs obtained from a database, and the target drug is one of multiple drugs obtained from a database.

[0079] The specific process of performing neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug in step 1 to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug is as follows:

[0080] Running the graph convolutional network method on the associated bipartite graph, by calculating the formula

[0081]

[0082]

[0083]

[0084] Get the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug in, represents the initial feature vector of the i-th mRNA to be tested output by the l-th layer neural network, represents the initial feature vector of the jth target drug output by the lth layer neural network, i=1,2,...,I, I represents the total number of mRNAs pre-acquired in step 1, j=1,2,...,J, J represents the total number of drugs pre-acquired in step 1, Represents the aggregated information of neighbor nodes of the mRNA to be tested in the associated bipartite graph, Represents the neighbor node aggregation information of the target drug in the associated bipartite graph, w represents the vector dimension, R represents the real number domain, σ(·) represents the nonlinear activation function, represents the normalized adjacency matrix, D (m) Represents the angle matrix of the mRNA to be tested, D (m) ∈R I×I , D (d) Denotes the diagonal matrix of the target drug, D (d) ∈R J×J , N i Represents the central node m of the mRNA to be tested in the associated bipartite graph i Neighboring nodes, N j Represents the central node d of the target drug in the associated bipartite graph j The neighboring nodes of Represents the vector representation of the i-th mRNA to be tested output by the l-1 layer of the graph convolutional neural network model, represents the vector representation of the j-th target drug output by the l-1 layer of the graph convolutional neural network model, represents the vector representation after the aggregation of the i-th mRNA information in the l-th layer neural network, A represents the vector representation of the j-th drug information aggregation in the l-th layer neural network, i,j Represents the adjacency matrix between the i-th mRNA to be tested and the j-th target drug.

[0085] It is worth mentioning that this application introduces residual connections in the process of local neighbor information aggregation, which can emphasize the central node (m i and d j ) and alleviate the over-smoothing problem in the aggregation process of the graph convolution network layer.

[0086] Step 2: Based on the association bipartite graph, an association hypergraph is constructed to obtain the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug.

[0087] It should be noted that in some embodiments of the present application, a hypergraph convolutional network is used to describe the high-order representation between the mRNA nodes to be tested (target drug nodes) in the associated bipartite graph to obtain a vector representation (hypergraph feature vector representation) of the mRNA nodes to be tested and the target drug nodes based on the hypergraph convolutional network.

[0088] The specific process of step 2 is: first, construct an associated hypergraph based on the associated bipartite graph. This process can be implemented using a common hypergraph construction method, which will not be described in detail here; then, this application uses a parameterized structure matrix (H (m) and H (d) ) to reduce the parameters of the model, thereby simplifying the calculation; then, the hypergraph information is aggregated. In the embodiment of the present application, a residual connection is introduced in the hypergraph information aggregation layer (hypergraph convolutional network information aggregation layer), which can improve the accuracy of solving the hypergraph vector representation, and finally obtain the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug.

[0089] For example, by calculating the formula

[0090]

[0091]

[0092]

[0093] ψ(X)=σ(VX)+X

[0094] Get the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug in, It represents the hyperedge feature vector representation of the mRNA to be tested at the lth hypergraph information aggregation layer, represents the hyperedge feature vector representation of the target drug at the lth hypergraph information aggregation layer, W (m) Represents the feature matrix of the hyperedge of the mRNA to be tested, W (m) ∈R d×H , W (d) The characteristic matrix of the hyperedge representing the target drug, W (d) ∈R d×H , Λ (m) Represents the hypergraph feature vector of the mRNA to be tested, Λ (d) Represents the hypergraph feature vector of the target drug, Λ (m) The vector representing the mRNA to be tested is E (m) The hypergraph structure matrix H corresponding to the mRNA to be tested (m) Polymerization, Λ (d) The vector representation E of the target drug(d) The hypergraph structure matrix H corresponding to the target drug (d) Aggregation is obtained, Λ∈R H×d , E∈R K×d , H∈R K×H , K=I or K=J, c represents the number of hypergraph vector representation layers, ψ(·) represents the encoding function of the hypergraph information aggregation layer, X represents the parameter matrix, X∈R H×H , σ' represents the LeakyReLU (leaky linear rectification) mapping.

[0095] In step 3, for each of the mRNA to be tested and the target drug, a cross-view contrast loss is constructed based on the initial feature vector and hypergraph feature vector representation of the mRNA to be tested and the target drug to be tested, and the cross-view contrast loss of the mRNA to be tested and the cross-view contrast loss of the target drug are obtained.

[0096] Specifically, based on the initial feature vector and hypergraph feature vector representation of the mRNA to be tested, the cross-view comparison loss of the mRNA to be tested is constructed; based on the initial feature vector and hypergraph feature vector representation of the target drug, the cross-view comparison loss of the target drug is constructed.

[0097] For example, by calculating the formula

[0098]

[0099] Get the view comparison loss of the mRNA to be tested and cross-view comparison loss of target drugs i- represents the negative sample of the i-th mRNA, and j- represents the negative sample of the j-th drug.

[0100] The negative samples above indicate that there is no verified association. For example, mRNA m,i- Indicates that there is no verified association between the mth mRNA and the i-th drug.

[0101] It should be noted that, in some embodiments of the present application, the same mRNA (z m,l ,Γ m,l ) and drugs (z d,l ,Γ d,l ) are taken as positive sample pairs, and different views of different mRNAs and drugs are taken as negative sample pairs.

[0102] In some embodiments of the present application, the formula

[0103]

[0104] Alleviate the overfitting problem introduced by the cross-view contrastive learning method; where := represents the assignment operator, Represents element-wise multiplication, M G ∈R I×J and M H ∈R J×H Both represent binary mask matrices with missing probability μ.

[0105] In step 4, a comprehensive loss is constructed based on the cross-view comparison loss of the mRNA to be tested and the cross-view comparison loss of the target drug.

[0106] Specifically, by calculating the formula

[0107]

[0108]

[0109] Get the comprehensive loss L; where, represents the association score of the positive sample, represents the association score of negative samples, Θ represents the model parameters, r i represents the i-th mRNA, m j represents the jth drug.

[0110] In some embodiments of the present application, the above-mentioned comprehensive loss L needs to be minimized using the Adam (Adaptive Moment Estimation) optimizer to achieve the effect of model fitting (the obtained eigenvector is the most accurate).

[0111] In step 5, the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug are back-propagated and updated respectively using the comprehensive loss to obtain the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug.

[0112] In some embodiments of the present application, the specific process of the above-mentioned back propagation update is to call the loss back propagation of pytorch (an open source Python machine learning library) to perform gradient update on the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug.

[0113] Step 6: If the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet the preset update termination criteria, the intermediate feature vector of the mRNA to be tested is used as the final feature vector of the mRNA to be tested, and the intermediate feature vector of the target drug is used as the final feature vector of the target drug; otherwise,

[0114] The intermediate feature vector of the mRNA to be tested is used as the vector representation of the mRNA to be tested in step 1, and the intermediate feature vector of the target drug is used as the vector representation of the target drug in step 1, and the process returns to step 1.

[0115] In an embodiment of the present application, the determination process of the above update termination criteria is as follows:

[0116] Step a: Calculate the AUC value and AUPR value based on the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug, respectively.

[0117] The AUC value (the area under the ROC curve (receiver operating characteristic curve) and the coordinate axis) and AUPR value (the area under the RC curve and the coordinate axis) are calculated here to determine whether the intermediate feature vector of the lncRNA to be tested and the intermediate feature vector of the target drug at this time have met the update termination condition (the optimal intermediate feature vector of the mRNA to be tested and the optimal intermediate feature vector of the target drug). If the update termination condition is not met, this step will be repeated until the model is fitted (the optimal intermediate feature vector of the mRNA to be tested and the optimal intermediate feature vector of the target drug are solved), which improves the accuracy of the feature vector and thus improves the accuracy of the association prediction.

[0118] It should be noted that the calculation of AUC and AUPR values is common knowledge, and the calculation process will not be described here in detail.

[0119] In step b, if the AUC value and the AUPR value reach the maximum value, it is determined that the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet the preset update termination condition; otherwise, it is determined that the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug do not meet the preset update termination condition.

[0120] Step 7: Construct a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug.

[0121] Step 8: Use the correlation prediction model to predict the correlation between the mRNA to be tested and the target drug.

[0122] Exemplarily, the final feature vector of the mRNA to be tested and the final feature vector of the target drug are input into the correlation prediction model to obtain the correlation score between the mRNA to be tested and the target drug. The larger the correlation score, the stronger the correlation between the mRNA to be tested and the target drug.

[0123] The specific process of step 7 (constructing a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug) is exemplified below.

[0124] Step 7.1, calculate the formula

[0125]

[0126] Get the final vector representation of the mRNA to be tested and the final vector representation of the target drug

[0127] in, represents the vector representation of the i-th mRNA information after aggregation in the l-th layer hypergraph convolutional network, Represents the vector representation after aggregation of the j-th drug information in the l-th layer hypergraph convolutional network.

[0128] Step 7.2, according to and Construct a correlation prediction model.

[0129] The expression of the correlation prediction model constructed in step 7.2 is as follows:

[0130]

[0131] in, represents the feature vector of the i-th mRNA in the l-th layer of the graph convolutional network, represents the feature vector of the jth drug in the lth layer of the graph convolutional network, represents the association score between the i-th mRNA and the j-th drug.

[0132] In some embodiments, in order to verify the accuracy of the method for predicting the association between mRNA and drugs provided in this application, the method for predicting the association between mRNA and drugs provided in this application was compared with the performance of the current existing technology, and the results were as follows: Figure 2 and Figure 3 As shown, Figure 2 The horizontal axis represents the false positive rate, Figure 2 The vertical axis represents the true rate, Figure 3 The horizontal axis represents the recall rate, Figure 3 The vertical axis represents the accuracy. Figure 2 and Figure 3), MF represents the matrix decomposition algorithm; GCN represents the graph convolutional network algorithm; LightGCN represents an optimization method for graph convolutional networks, which abandons the feature changes and nonlinear activation of traditional graph convolutional networks and only retains the node aggregation of graph convolutional networks; GCL-ED represents the contrastive learning method of the data augmentation method based on random loss of edges; GCL-ND represents the contrastive learning method of the data augmentation method based on random loss of nodes; GCL-RW represents the contrastive learning method of the data augmentation method based on random walks; MLRDFM represents the method of MLRDFM integrating the similarities of four miRNAs and two diseases, and predicting the association between miRNAs and diseases through a deep decomposition machine method; HCCFMDA represents the prediction method for the association between mRNA and drugs provided in this application. Comprehensive comparison shows that the performance of the prediction method for the association between mRNA and drugs provided in this application is better than other existing technologies.

[0133] In other embodiments, in order to verify the impact of graph convolutional networks, hypergraph convolutional networks, and contrastive learning on the performance of the present invention, different experiments were conducted, and the experimental results are as follows:

[0134]

[0135]

[0136] P means the component is used, and O means the component is not used.

[0137] As can be seen from the above table, if only the graph convolutional network is used to aggregate neighbor node information or only the hypergraph convolutional network is used to obtain high-order information between nodes, the effect of using the hypergraph convolutional network is more significant, and the combination of the two also significantly improves the prediction performance of the model.

[0138] It can be seen from the above steps that the method for predicting the association between mRNA and drugs provided in the present application constructs a cross-view comparison loss based on the initial feature vector and hypergraph feature vector representation of the mRNA to be tested and the target drug, and then backpropagates and updates the initial feature vectors of the mRNA to be tested and the target drug respectively according to the cross-view comparison loss, so as to obtain a more accurate feature vector, thereby improving the accuracy of the prediction of the association between the mRNA to be tested and the target drug.

[0139] The following is an illustrative description of the device for predicting the association between mRNA and drugs provided in this application with reference to specific examples.

[0140] like Figure 4 As shown, an embodiment of the present application provides a device for predicting the association between mRNA and drugs, and the device 400 for predicting the association between mRNA and drugs includes:

[0141] Initialization module 401 is used to run an information aggregation neural network model on a pre-constructed bipartite graph of association between the mRNA to be tested and the target drug, and perform neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug, respectively, to obtain an initial feature vector of the mRNA to be tested and an initial feature vector of the target drug;

[0142] A hypergraph feature vector module 402 is used to construct an association hypergraph based on the association bipartite graph to obtain a hypergraph feature vector representation of the mRNA to be tested and a hypergraph feature vector representation of the target drug;

[0143] Contrast loss module 403 is used to construct a cross-view contrast loss for either the test mRNA or the target drug based on the initial feature vector and hypergraph feature vector representation of the test mRNA and the target drug, respectively, to obtain the cross-view contrast loss of the test mRNA and the cross-view contrast loss of the target drug;

[0144] A feature vector optimization module 404 is configured to construct a comprehensive loss based on the cross-view comparison loss function of the mRNA to be tested and the cross-view comparison loss function of the target drug;

[0145] The intermediate feature vector module 405 is used to perform backpropagation updates on the initial feature vectors of the mRNA to be tested and the initial feature vectors of the target drug using the comprehensive loss to obtain the intermediate feature vectors of the mRNA to be tested and the intermediate feature vectors of the target drug;

[0146] The final feature vector module 406 is used to use the intermediate feature vector of the mRNA to be tested as the final feature vector of the mRNA to be tested and the intermediate feature vector of the target drug as the final feature vector of the target drug if the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet the preset update termination criteria; otherwise,

[0147] The intermediate feature vector of the mRNA to be tested is used as the vector representation of the mRNA to be tested in the initialization module, and the intermediate feature vector of the target drug is used as the vector representation of the target drug in the initialization module, and the initialization module is returned to execute;

[0148] The correlation prediction model module 407 is used to construct a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug;

[0149] The prediction module 408 is used to predict the correlation between the mRNA to be tested and the target drug using the correlation prediction model.

[0150] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment section and will not be repeated here.

[0151] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0152] like Figure 5 As shown, an embodiment of the present application provides a terminal device, such as Figure 5 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 5 Only one processor is shown in the figure), a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 implements the steps of any of the above method embodiments when executing the computer program D102.

[0153] Specifically, when the processor D100 executes the computer program D102, the information aggregation neural network model is run on the pre-constructed association bipartite graph of the mRNA to be tested and the target drug, and the neighbor node aggregation is performed on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug, respectively, to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug, and then, based on the association bipartite graph, an association hypergraph is constructed to obtain the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug, and then, for each of the mRNA to be tested and the target drug, a cross-view comparison loss is constructed based on the initial feature vector and the hypergraph feature vector representation of the mRNA to be tested and the target drug, to obtain the view comparison loss of the mRNA to be tested and the cross-view comparison loss of the target drug, and then, based on the mRNA to be tested and the target drug, a cross-view comparison loss is constructed based on the initial feature vector and the hypergraph feature vector representation of the mRNA to be tested and the target drug, respectively. The cross-view contrast loss of the NA and the cross-view contrast loss of the target drug are used to construct a comprehensive loss. Subsequently, the comprehensive loss is used to backpropagate and update the initial feature vectors of the test mRNA and the initial feature vectors of the target drug, respectively, to obtain the intermediate feature vectors of the test mRNA and the intermediate feature vectors of the target drug. If the intermediate feature vectors of the test mRNA and the intermediate feature vectors of the target drug meet the preset update termination criteria, the intermediate feature vector of the test mRNA is used as the final feature vector of the test mRNA, and the intermediate feature vector of the target drug is used as the final feature vector of the target drug. Then, based on the final feature vectors of the test mRNA and the final feature vectors of the target drug, a correlation prediction model between the test mRNA and the target drug is constructed. Finally, the correlation prediction model is used to predict the correlation between the test mRNA and the target drug. Among them, the cross-view contrast loss is constructed based on the initial feature vectors and hypergraph feature vector representations of the test mRNA and the target drug, and the initial feature vectors of the test mRNA and the target drug are backpropagated and updated based on the cross-view contrast loss. More accurate feature vectors can be obtained, thereby improving the accuracy of the prediction of the correlation between the test mRNA and the target drug.

[0154] The processor D100 may be a central processing unit (CPU), or may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0155] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may also be an external storage device of the terminal device D10, such as a plug-in hard disk, a smart memory card (SMC, SmartMedia Card), a secure digital (SD, Secure Digital) card, a flash card, etc. equipped on the terminal device D10. Furthermore, the memory D101 may also include both an internal storage unit of the terminal device D10 and an external storage device. The memory D101 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory D101 may also be used to temporarily store data that has been output or is to be output.

[0156] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned various method embodiments can be implemented.

[0157] An embodiment of the present application provides a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0158] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device capable of carrying the computer program code to the prediction device / terminal device associated with mRNA and drug, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. For example, USB flash drive, mobile hard disk, magnetic disk or optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signal and telecommunication signal.

[0159] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0160] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0161] In the embodiments provided in this application, it should be understood that the disclosed devices / network equipment and methods can be implemented in other ways. For example, the device / network equipment embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0162] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0163] This application has the following advantages:

[0164] 1. We propose a general computational framework for predicting the association between mRNA and drug sensitivity. In this framework, we use a graph convolutional network method to aggregate high-order relationships between neighbor nodes on an mRNA-drug bipartite graph. We also use a hypergraph convolutional network method to obtain high-order relationships between nodes in the mRNA-drug bipartite graph. We also introduce a hypergraph contrastive learning method to improve the predictive ability of the model. Finally, we use the inner product to obtain the association score of a specific mRNA-drug sensitivity. A higher score indicates that the mRNA and drug are more likely to have a potential association.

[0165] 2. Large-scale prediction of mRNA-drug sensitivity association pairs can be carried out in a very short time, effectively solving the problems of high cost and high blindness of biological experiments.

[0166] The above is a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for predicting the association between mRNA and drugs, characterized in that: include: Step 1: running an information aggregation neural network model on a pre-constructed association bipartite graph consisting of the mRNA to be tested and the target drug, performing neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug, respectively, to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug; Step 2: constructing an association hypergraph based on the association bipartite graph to obtain a hypergraph feature vector representation of the mRNA to be tested and a hypergraph feature vector representation of the target drug; Step 3: construct a cross-view contrast loss for each of the mRNA to be tested and the target drug based on the initial feature vector and hypergraph feature vector representation of the mRNA to be tested and the target drug to be tested, respectively. Step 4, constructing a comprehensive loss based on the cross-view comparison loss of the mRNA to be tested and the cross-view comparison loss of the target drug; Step 5, using the comprehensive loss, back-propagating and updating the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug, respectively, to obtain the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug; Step 6: If the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet the preset update termination criteria, the intermediate feature vector of the mRNA to be tested is used as the final feature vector of the mRNA to be tested, and the intermediate feature vector of the target drug is used as the final feature vector of the target drug; otherwise, Using the intermediate feature vector of the mRNA to be tested as the vector representation of the mRNA to be tested in step 1, and the intermediate feature vector of the target drug as the vector representation of the target drug in step 1, and returning to step 1; Step 7, constructing a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug; Step 8: Use the correlation prediction model to predict the correlation between the mRNA to be tested and the target drug.

2. The prediction method according to claim 1, characterized in that The information aggregation neural network model in step 1 is a graph convolutional neural network model; The step 1 of performing neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug includes: By calculating the formula Get the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug in, represents the initial feature vector of the i-th mRNA to be tested output by the l-th layer neural network, represents the initial feature vector of the jth target drug output by the lth layer neural network, i = 1, 2, ..., I, I represents the total number of mRNAs pre-acquired in step 1, j = 1, 2, ..., J, J represents the total number of drugs pre-acquired in step 1, represents the aggregation information of neighbor nodes of the mRNA to be tested in the associated bipartite graph, represents the neighbor node aggregation information of the target drug in the association bipartite graph, w represents the vector dimension, R represents the real number domain, σ(·) represents the nonlinear activation function, represents the normalized adjacency matrix, D (m) Denotes the angle matrix of the mRNA to be tested, D (m) ∈R I×I , D (d) Denotes the diagonal matrix of the target drug, D (d) ∈R J×J , N i Represents the central node m of the mRNA to be tested in the association bipartite graph i Neighboring nodes, N j Represents the central node d of the target drug in the association bipartite graph j The neighboring nodes of represents the vector representation of the i-th mRNA to be tested output by the l-1 layer of the graph convolutional neural network model, represents the vector representation of the j-th target drug output by the l-1 layer of the graph convolutional neural network model, represents the vector representation after the aggregation of the i-th mRNA information in the l-th layer neural network, A represents the vector representation of the j-th drug information aggregation in the l-th layer neural network, i,j Represents the adjacency matrix between the i-th mRNA to be tested and the j-th target drug.

3. The prediction method according to claim 2, characterized in that The step 2 includes: By calculating the formula ψ(X)=σ(VX)+X Obtain the hypergraph feature vector representation of the mRNA to be tested and the hypergraph feature vector representation of the target drug in, represents the hyperedge feature vector representation of the mRNA to be tested at the lth hypergraph information aggregation layer, represents the hyperedge feature vector representation of the target drug at the lth hypergraph information aggregation layer, W (m) Represents the feature matrix of the hyperedge of the mRNA to be tested, W (m) ∈R d×H , W (d) The characteristic matrix of the hyperedge representing the target drug, W (d) ∈R d×H , Λ (m) Represents the hypergraph feature vector of the mRNA to be tested, Λ (d) Represents the hypergraph feature vector of the target drug, Λ (m) The vector representing the mRNA to be tested is E (m) and the hypergraph structure matrix H corresponding to the mRNA to be tested (m) Polymerization, Λ (d) The target drug is represented by a vector E (d) The hypergraph structure matrix H corresponding to the target drug (d) Aggregation is obtained, Λ∈R H×d , E∈R K ×d , H∈R K×H , K=I or K=J, c represents the number of hypergraph vector representation layers, ψ(·) represents the encoding function of the hypergraph information aggregation layer, X represents the parameter matrix, X∈R H×H , σ' represents the LeakyReLU mapping.

4. The prediction method according to claim 3, characterized in that The step 3 includes: By calculating the formula Obtain the view comparison loss of the mRNA to be tested and the cross-view contrast loss of the target drug i- represents the negative sample of the i-th mRNA, and i- represents the negative sample of the i-th drug.

5. The prediction method according to claim 4, characterized in that The step 4 comprises: By calculating the formula The comprehensive loss L is obtained; wherein, represents the association score of the positive sample, represents the association score of negative samples, Θ represents the model parameters, r i represents the i-th mRNA, m j represents the jth drug.

6. The prediction method according to claim 5, characterized in that The step 7 comprises: By calculating the formula Get the final vector representation of the mRNA to be tested and the final vector representation of the target drug in, represents the vector representation of the i-th mRNA information after aggregation in the l-th layer hypergraph convolutional network, Represents the vector representation of the j-th drug information after aggregation in the l-th layer hypergraph convolutional network; according to and Constructing the correlation prediction model.

7. The prediction method according to claim 6, characterized in that The expression of the correlation prediction model is as follows: in, represents the feature vector of the i-th mRNA in the l-th layer of the graph convolutional network, represents the feature vector of the jth drug in the lth layer of the graph convolutional network, represents the association score between the i-th mRNA and the j-th drug.

8. A prediction device for the association between mRNA and drugs, characterized in that: include: An initialization module is used to run an information aggregation neural network model on a pre-constructed bipartite graph of association between the mRNA to be tested and the target drug, and perform neighbor node aggregation on the pre-initialized vector representation of the mRNA to be tested and the pre-initialized vector representation of the target drug, respectively, to obtain the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug; A hypergraph feature vector module is used to construct an association hypergraph based on the association bipartite graph to obtain a hypergraph feature vector representation of the mRNA to be tested and a hypergraph feature vector representation of the target drug; a contrast loss module for constructing a cross-view contrast loss for each of the test mRNA and the target drug based on the initial feature vector and the hypergraph feature vector representation of the test mRNA and the target drug, respectively, to obtain the view contrast loss of the test mRNA and the cross-view contrast loss of the target drug; A feature vector optimization module, configured to construct a comprehensive loss based on the cross-view comparison loss function of the mRNA to be tested and the cross-view comparison loss function of the target drug; An intermediate feature vector is used to back-propagate and update the initial feature vector of the mRNA to be tested and the initial feature vector of the target drug using the comprehensive loss to obtain an intermediate feature vector of the mRNA to be tested and an intermediate feature vector of the target drug; a final feature vector module, configured to use the intermediate feature vector of the mRNA to be tested as the final feature vector of the mRNA to be tested, and the intermediate feature vector of the target drug as the final feature vector of the target drug, if the intermediate feature vector of the mRNA to be tested and the intermediate feature vector of the target drug meet a preset update termination criterion; otherwise, Using the intermediate feature vector of the mRNA to be tested as the vector representation of the mRNA to be tested in the initialization module, and the intermediate feature vector of the target drug as the vector representation of the target drug in the initialization module, and returning to execute the initialization module; A correlation prediction model module is used to construct a correlation prediction model between the mRNA to be tested and the target drug based on the final feature vector of the mRNA to be tested and the final feature vector of the target drug; The prediction module is used to predict the correlation between the mRNA to be tested and the target drug using the correlation prediction model.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for predicting the association between mRNA and drugs according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for predicting the association between mRNA and drugs according to any one of claims 1 to 7 is implemented.

Images