IncRNA-disease association prediction method based on KAN network

Abstract

The invention is suitable for the field of lncRNA and disease association prediction, and particularly provides an lncRNA-disease association prediction method based on a KAN network, and the method comprises the following steps: fusing lncRNA-disease association network information and various similarity information of lncRNA and diseases to construct a comprehensive heterogeneous network; the method comprises the following steps of: constructing a Transform feature extractor based on a multi-head attention mechanism, wherein the Transform feature extractor is used for capturing global serialization features from lncRNA and a disease node; constructing a GraphSAGE feature extractor based on a heterogeneous graph network, wherein the GraphSAGE feature extractor is used for capturing topological structure features of lncRNA and a disease node; the global serialization features and the topological structure features are fused, and comprehensive lncRNA-disease association pair feature information is obtained; and a KAN network used for predicting the association score between the lncRNA and the disease is constructed. According to the method, global sequence features are extracted in combination with Transform, heterogeneous network topology information is mined through GraphSAGE, a nonlinear relation is modeled through a KAN network, multilevel information of lncRNA and diseases is fully mined, and the accuracy of correlation prediction between lncRNA and diseases is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of lncRNA and disease association prediction, and in particular relates to a lncRNA-disease association prediction method based on a KAN network. Background Art

[0002] lncRNA is a class of non-coding RNA molecules that do not encode proteins but have important regulatory functions. They are widely involved in gene transcription, epigenetic regulation and RNA interactions. They play a key role in cancer occurrence, development and drug resistance. lncRNA can also affect cancer-related signaling pathways and be used as a biomarker for cancer diagnosis, prognosis assessment and targeted therapy, and has broad application prospects.

[0003] In related technologies, for example, LINC01608 has been experimentally determined to be a reliable prognostic biomarker for hepatocellular carcinoma; overexpression of lncRNACTA-929C8 in brain tissue may contribute to the development of Alzheimer's disease. Therefore, accurate identification of disease-related lncRNA is of great significance for the diagnosis of the disease and the revelation of its pathogenesis. In addition, for example, WMFLDA designed a weighted matrix decomposition method to infer disease-related lncRNAs, thereby predicting the association information between them; MCA-Net proposed a multi-feature encoding method, combining 6 similarity features to construct the association features between lncRNA and disease, and using convolutional neural networks to infer the potential association between lncRNA and disease. Therefore, the method based on graph neural network has become an effective way to solve the problem of lncRNA-disease association prediction. However, the model based on a single graph neural network has limited performance due to excessive smoothing.

[0004] At present, traditional lncRNA-disease association prediction methods mainly include experimental and computational biology-based methods. Experimental-based methods require expensive and time-consuming experimental processes to measure the association between lncRNA and disease, while computational biology-based methods use known lncRNA and disease data for prediction. However, these traditional methods have problems such as incomplete data, limited feature representation, and limitations of prediction models; moreover, the availability and quality of data are limited, the feature representation lacks comprehensiveness, and the prediction model cannot fully consider the complex association between lncRNA and disease. Therefore, the existing prediction methods have the following problems: First, existing methods mainly extract node features based on lncRNA sequence similarity and disease semantic similarity, but do not fully exploit the multi-source feature information of lncRNA and disease, limiting the model's ability to depict complex biological relationships; Second, the feature extraction methods of existing methods cannot effectively extract information about lncRNA and diseases. For example, convolutional neural networks are mainly good at local feature extraction and cannot effectively capture long-distance dependency information in heterogeneous networks. Summary of the invention

[0005] The purpose of the embodiment of the present invention is to provide a lncRNA-disease association prediction method based on the KAN network, which combines Transformer to extract global sequence features and GraphSAGE to mine heterogeneous network topology information, and models nonlinear relationships through the KAN network to fully mine the multi-level information of lncRNA and diseases and improve the accuracy of association prediction.

[0006] To achieve the above object, the present invention provides the following technical solutions: A lncRNA-disease association prediction method based on a KAN network comprises the following steps: S1, integrating lncRNA-disease association network information and multiple similarity information of lncRNA and disease to construct a comprehensive heterogeneous network; S2. Construct a Transformer feature extractor based on a multi-head attention mechanism to capture global serialization features from lncRNA and disease nodes; S3. Build a GraphSAGE feature extractor based on heterogeneous graph networks to capture the topological structural features of lncRNA and disease nodes. S4. The global serialization features and topological structure features are integrated to obtain the comprehensive lncRNA-disease association pair feature information. A KAN network is constructed to predict the association score between lncRNA and disease. The KAN network includes a B-Spline transformation layer and a deep KAN structure. The learnable B-spline interpolation of the transformation layer is used to learn the nonlinear mapping relationship of the lncRNA-disease association pair feature information. The deep KAN structure is used to predict the association score between lncRNA and disease.

[0007] Furthermore, in step S1, an m×n association network is used to represent the association information of lncRNA-disease, where m represents the number of lncRNAs and n represents the number of diseases. i and disease j When there is a correlation, the corresponding position of A (l i , d j ) is 1, otherwise it is 0; The association network definition is as follows: ; Calculate the preliminary characteristic information of lncRNA and disease, including lncRNA sequence similarity lS seq , lncRNA functional similarity lS func and lncRNA Gaussian interaction spectrum kernel similarity lS GIPK , Disease semantic similarity DS sem and disease Gaussian interaction spectrum kernel similarity DS GIPK ; The preliminary feature information of lncRNA and disease were fused respectively to obtain the lncRNA comprehensive similarity feature LS and the disease comprehensive similarity feature DS, where: , ; Finally, the lncRNA-disease association network information, the comprehensive similarity feature lS, and the disease comprehensive similarity feature DS were integrated to construct a lncRNA-disease heterogeneous network: , where lS and DS represent the comprehensive similarity feature matrix of lncRNA and the comprehensive similarity feature matrix of diseases, respectively, and A represents their adjacency matrix. T Represents the transpose of the adjacency matrix.

[0008] Furthermore, lncRNA sequence similarity lS seq In the calculation process, the Smith-Waterman sequence alignment algorithm was used to evaluate the sequence similarity between lncRNAs; the formula for calculating sequence similarity is as follows: ; Among them, l i and l j denote the i-th lncRNA and the j-th lncRNA, respectively, SW(l i , l j ) indicates that the sequence alignment score is calculated based on the Smith-Waterman alignment algorithm.

[0009] Furthermore, lncRNA functional similarity lS func In the calculation process, the formula is expressed as: ; Among them, p and q represent the 1 and l 2 The number of associated diseases, de represents a disease, DE represents a group of diseases, DE={de 1 , 2 , ...}, S(de, DE) is used to calculate the relationship between diseases, expressed as: .

[0010] Furthermore, the lncRNA Gaussian interaction spectrum kernel similarity lS GIPK In the calculation process, the GIPK nuclear similarity calculation formula between lncRNAs is as follows: ; Among them, A(l i ,) and A(l j ,) represent the i-th row vector and j-th row vector of A respectively; λ n is the kernel width coefficient, defined as: , where N represents the total number of lncRNAs, A(l k ,) represents the k-th row vector of A.

[0011] Furthermore, disease semantic similarity DS sem In the calculation process, the disease ontology DO is used to measure the semantic similarity between related diseases, which is represented as a directed acyclic graph. By utilizing the hierarchical structure of diseases in DO, the semantic similarity is calculated based on the directed acyclic graph. The calculation formula is expressed as: ; Among them, d i and d j represent the i-th disease and the j-th disease respectively, T i and T j denotes a set of sets containing the i-th and j-th diseases, respectively, S di (t) represents the semantic impact of disease t on the i-th disease.

[0012] Furthermore, the disease Gaussian interaction spectrum kernel similarity DS GIPK In the calculation process, the calculation formula is expressed as: ; Among them, A(,d i ) and A(,d j ) represent the i-th column vector and j-th column vector of A respectively, λ d It is expressed as: , D represents the total number of diseases, A(,n l ) represents the lth column vector of A.

[0013] Furthermore, in step S2, the Transformer feature extractor includes three encoding layers, and in the hth attention head of one of the encoding layers, three linear transformation matrices are included: the query matrix 、key matrix and value matrix , the specific linear transformation is: ; in, Represents the weight matrix of the query matrix; Represents the weight matrix of the key matrix; Represents the weight matrix of the value matrix; H l-1 Represents the output of the l-1th layer; Calculate the attention score matrix , the output of each layer is obtained by normalizing the attention score matrix and the value matrix. The specific formula is expressed as: ; Where d represents and The matrix dimensions of represents the output of the lth layer in the hth attention head; The final result of each layer is obtained by concatenating the results of each attention head by multiple attention heads; after the output of the three-layer Transformer encoding layer, the feature representation F is extracted. 1 .

[0014] Furthermore, in step S3, the GraphSAGE feature extractor includes three layers of GraphSAGE computing layers, and feature extraction is achieved through neighbor sampling and aggregation mechanisms; wherein, for each lncRNA and disease node, its neighbor node set is randomly sampled, and the mean of its features is calculated, expressed as: ; Among them, W k is the weight matrix, u is the set of adjacent nodes of node v represented by N(v), MEAN represents the average operation, σ is the nonlinear activation function, and RELU is used; It is the node representation after calculation of the kth layer; By using multi-layer calculations, the output of each layer is used as the input of the next layer, and finally the representation feature matrix F of the heterogeneous network is obtained 2 .

[0015] Furthermore, in the step of fusing the global sequence features and the topological structure features to obtain the comprehensive lncRNA-disease association pair feature information, the fusion method is expressed as: ; Among them, the coefficient α 1 , α 2 Indicates the importance of the feature; The calculation method of the B-Spline transformation layer is expressed as: *=W·f(x)+b; where f(x) represents the feature after B-Spline interpolation, and W and b are trainable weights and bias terms; The loss function is expressed as follows: ; Among them, y i represents the true label of the lncRNA-disease association pair, represents the predicted label of the lncRNA-disease association pair, and N represents the number of lncRNA-disease association pairs.

[0016] Compared with the prior art, the technical advantages of the lncRNA-disease association prediction method based on the KAN network of the present invention are manifested in the following aspects: First, the lncRNA-disease association prediction method proposed in the present invention is used to identify potential lncRNA-disease associations. This method combines Transformer to extract global sequence features, GraphSAGE to mine heterogeneous network topology information, and KAN network to model nonlinear relationships, fully mining the multi-level information of lncRNA and diseases, and improving the accuracy of association prediction; Second, the present invention combines two feature learning extractors, Transformer and GraphSAGE, to capture the global feature information of lncRNA and disease and the network topology information respectively; the two modules cooperate with each other, enabling TransSAGE-KAN-LDA to more accurately characterize the complex relationship between lncRNA and disease, and improve the accuracy and generalization ability of prediction.

[0017] In summary, the present invention combines two feature learning modules, Transformer and GraphSAGE, to extract the global feature information of lncRNA and disease and capture the topological structure information of heterogeneous networks respectively; uses the KAN network for feature association prediction, fully integrating multi-source information to more accurately identify the potential association between lncRNA and disease; through the complementary learning of Transformer and GraphSAGE, this method can effectively characterize the complex relationship between lncRNA and disease, and improve the prediction ability and generalization performance of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0019] Figure 1It is a schematic diagram of the implementation process of the lncRNA-disease association prediction method based on the KAN network of the present invention; Figure 2 It is a schematic diagram of the ROC curve on the data set under the five-fold cross validation of the present invention; Figure 3 It is a schematic diagram of the PR curve on the data set under the five-fold cross validation of the present invention; Figure 4 This is a system architecture diagram of the lncRNA-disease association prediction method based on the KAN network of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0022] Please refer to Figure 1 and Figure 4 In one embodiment provided by the present invention, a method for predicting lncRNA-disease association based on a KAN network is provided, the method comprising the following steps: S1. Integrate lncRNA-disease association network information and multiple similarity information of lncRNA and disease to build a comprehensive heterogeneous network. In this step, data collection is required first, including lncRNA sequence data, disease DOID data and known lncRNA-disease association information in public databases. S2. Construct a Transformer feature extractor based on a multi-head attention mechanism to capture global serialization features from lncRNA and disease nodes; S3. Build a GraphSAGE feature extractor based on heterogeneous graph networks to capture the topological structural features of lncRNA and disease nodes. S4. The global serialization features and topological structure features are integrated to obtain the comprehensive lncRNA-disease association pair feature information. A KAN network is constructed to predict the association score between lncRNA and disease. The KAN network includes a B-Spline transformation layer and a deep KAN structure. The learnable B-spline interpolation of the transformation layer is used to learn the nonlinear mapping relationship of the lncRNA-disease association pair feature information. The deep KAN structure is used to predict the association score between lncRNA and disease.

[0023] Therefore, the association prediction method provided by the present invention combines two feature learning modules, Transformer and GraphSAGE, to extract the global feature information of lncRNA and disease and capture the topological structure information of heterogeneous networks respectively; further, the KAN network is used to perform feature association prediction, fully integrating multi-source information to more accurately identify the potential association between lncRNA and disease; Through the complementary learning of Transformer and GraphSAGE, the method can effectively characterize the complex relationship between lncRNA and disease, and improve the prediction ability and generalization performance of the model.

[0024] In addition, the association prediction method provided by the present invention also optimizes model parameters through back propagation and gradient descent algorithms.

[0025] In step S1 of the present invention, lncRNA sequence data, disease DOID data and known lncRNA-disease association information in public databases are collected; Among them, during the data collection process, RNADiseasev4.0 contained 3,428,058 lncRNA-disease associations, covering 17,820 lncRNAs and 4,090 diseases; Lnc2Cancer3.041 included a total of 1,057 associations, involving 531 lncRNAs and 86 human cancers; LncRNADiseasev3.0 collected 25,440 experimentally supported lncRNA-disease associations, covering 6,066 lncRNAs and 566 diseases; After merging the above three datasets, the lncRNA data were simply screened and duplicates were removed, resulting in 886 lncRNA-disease associations involving 406 lncRNAs and 255 diseases.

[0026] Then, in step S1 of the present invention, an m×n association network A∈R is used. m×n To represent the association information of lncRNA-disease, m represents the number of lncRNAs, n represents the number of diseases, and when lncRNA l i and disease j When there is a correlation, the corresponding position of A (l i , d j ) is 1, otherwise it is 0; where the association network is defined as follows: ; Calculate the preliminary characteristic information of lncRNA and disease, including lncRNA sequence similarity lS seq , lncRNA functional similarity lSfunc and lncRNA Gaussian interaction spectrum kernel similarity lS GIPK , Disease semantic similarity DS sem and disease Gaussian interaction spectrum kernel similarity DS GIPK ; The preliminary feature information of lncRNA and disease were fused respectively to obtain the lncRNA comprehensive similarity feature LS and the disease comprehensive similarity feature DS, where: , ; Finally, the lncRNA-disease association network information, the comprehensive similarity feature lS, and the disease comprehensive similarity feature DS were integrated to construct a lncRNA-disease heterogeneous network: , where lS and DS represent the comprehensive similarity feature matrix of lncRNA and the comprehensive similarity feature matrix of diseases, respectively, and A represents their adjacency matrix. T represents the transpose of the adjacency matrix; In addition, the present invention also calculates 5 types of similarity information related to lncRNA and diseases; the 5 types of similarity information include: lncRNA sequence similarity, lncRNA functional similarity and lncRNA Gaussian interaction spectrum kernel similarity, disease semantic similarity and disease Gaussian interaction spectrum kernel similarity; Among them, in one implementation, the lncRNA sequence similarity 1S seq In the calculation process of lncRNA sequence similarity, according to the principle that lncRNAs with similar sequences are more likely to have similar functions, the present invention uses the Smith-Waterman sequence alignment algorithm to evaluate the sequence similarity between lncRNAs; the formula for calculating sequence similarity is as follows: ; Among them, l i and l j Represent the i-th lncRNA and the j-th lncRNA respectively; SW(l i , l j ) represents lncRNA l calculated based on Smith-Waterman alignment algorithm i and lncRNA j The sequence alignment score between .

[0027] Among them, in one implementation, the lncRNA functional similarity 1S funcIn the calculation process, for lncRNA functional similarity, this method uses a direct method to reflect their functional similarity, aiming to reveal their potential association with diseases with similar pathological phenomena or symptoms; lncRNA functional similarity can be calculated and measured by evaluating the similarity between the diseases related to it, and the formula is expressed as: ; Among them, p and q represent the 1 and lncRNA 2 The number of associated diseases, de represents a disease, DE is defined as a group of diseases, DE={de 1 , 2 , ...}, S(de, DE) is used to calculate the relationship between diseases, expressed as: .

[0028] Among them, in one implementation, the lncRNA Gaussian interaction spectrum kernel similarity lS GIPK In the calculation process of , for lncRNA Gaussian interaction spectrum kernel similarity, Gaussian interaction feature kernel similarity is usually used to evaluate the similarity between two similar nodes in non-coding RNA-disease association prediction tasks, indicating that in a specific disease, lncRNAs with similar functions have similar patterns. The GIPK kernel similarity calculation formula between lncRNAs is as follows: ; Among them, A(l i ,) and A(l j ,) represent the i-th row vector and j-th row vector of A respectively; λ n is the kernel width coefficient, defined as: , where N represents the total number of lncRNAs, A(l k ,) represents the k-th row vector of A.

[0029] Furthermore, in one implementation, the disease semantic similarity DS sem In the calculation process of disease semantic similarity, the present invention uses Disease Ontology (DO) to measure the semantic similarity between related diseases, which is usually represented as a directed acyclic graph. By utilizing the hierarchical structure of diseases in DO and calculating their semantic similarity based on the directed acyclic graph, the more parent diseases shared by diseases, the greater the similarity between them. The calculation formula is expressed as: ; Among them, d i and d j represent the i-th disease and the j-th disease respectively, Ti and T j denotes a set of sets containing the i-th and j-th diseases, respectively, S di (t) represents the semantic impact of disease t on the i-th disease.

[0030] Among them, in one implementation, the disease Gaussian interaction spectrum kernel similarity DS GIPK In the calculation process, for the disease Gaussian interaction spectrum kernel similarity, similar to the lncRNA Gaussian interaction spectrum kernel similarity, only the formula is given, and the calculation formula is expressed as: ; Where A(,d i ) and A(,d j ) represent the i-th column vector and j-th column vector of A respectively, λ d is defined as: , D represents the total number of diseases, A(,n l ) represents the lth column vector of A.

[0031] Then, the present invention obtains a comprehensive lncRNA similarity network and a comprehensive disease similarity network by fusing the similarity networks of lncRNA and disease respectively.

[0032] Furthermore, in step S2, the present invention designs a Transformer feature extractor based on a multi-head attention mechanism, aiming to capture global serialization features from lncRNA and disease nodes. By introducing a multi-head attention mechanism, the model is allowed to learn multiple different representations in parallel, reducing the dependence on a single representation, making the learning process more robust. Specifically, the Transformer feature extractor provided by the present invention includes three encoding layers, in which the hth attention head of one encoding layer contains three linear transformation matrices: query matrix 、key matrix and value matrix , the specific linear transformation is: ; in, Represents the weight matrix of the query matrix; Represents the weight matrix of the key matrix; Represents the weight matrix of the value matrix; H l-1 Represents the output of the l-1th layer; Calculate the attention score matrix , the output of each layer is obtained by normalizing the attention score matrix and the value matrix. The specific formula is expressed as: ; Where d represents and The matrix dimensions of represents the output of the lth layer in the hth attention head; The final result of each layer is obtained by concatenating the results of each attention head by multiple attention heads; after the output of the three-layer Transformer encoding layer, the feature representation F is extracted. 1 .

[0033] Further, in step S3, the present invention designs a GraphSAGE feature extractor based on a heterogeneous graph network, aiming to capture the topological structure information of the heterogeneous network. The GraphSAGE feature extractor provided by the present invention comprises three layers of GraphSAGE calculation layers, and realizes feature extraction through neighbor sampling and aggregation mechanism; wherein, for each lncRNA and disease node, its neighbor node set is randomly sampled, and the mean of its feature is calculated, which is expressed as: ; Among them, W k is the weight matrix, u is the set of adjacent nodes of node v represented by N(v), MEAN represents the average operation, σ is the nonlinear activation function, and preferably, the activation function σ adopts the RELU function; is the node representation after calculation at the kth layer; this process can effectively aggregate the local information of lncRNA and disease nodes, so that the final node representation can integrate the characteristics of neighbors; by adopting multi-layer calculation, the output of each layer is used as the input of the next layer, and finally the representation feature matrix F of the heterogeneous network is obtained 2 .

[0034] Furthermore, in the step of fusing the global sequence features and the topological structure features to obtain the comprehensive lncRNA-disease association pair feature information, the fusion method is expressed as: ; Among them, the coefficient α 1 , α 2 Indicates the importance of the feature; In one implementation of the present invention, a new network KAN (Kolmogorov-Arnold Network) is constructed, whose core components include a B-Spline transformation layer and a deep KAN structure. The nonlinear mapping relationship of the data is learned by using learnable B-spline interpolation, avoiding the limitation of fixed activation functions and improving parameter efficiency. Specifically, the calculation method of the B-Spline transformation layer is expressed as: *=W·f(x)+b; Among them, f(x) represents the feature after B-Spline interpolation, W and b are trainable weights and bias terms; Furthermore, the loss function is expressed as follows: ; Among them, y i represents the true label of the lncRNA-disease association pair, represents the predicted label of the lncRNA-disease association pair, and N represents the number of lncRNA-disease association pairs.

[0035] For further information, please refer to Figure 3 and Figure 4 In verifying the effectiveness of the embodiment of the present invention, the present invention uses five-fold cross validation to evaluate the performance of the model, which helps to comprehensively evaluate the generalization ability of the model and can effectively prevent the model from overfitting.

[0036] Specifically, the present invention regards all known lncRNA disease associations as positive samples, and unknown lncRNA disease associations mean that there is no relevant biological experiment to verify their correlation. Therefore, all unknown lncRNA-disease associations are regarded as negative samples. To ensure the balance of positive and negative samples, the present invention randomly selects the same number of negative samples as positive samples from all negative samples.

[0037] Since AUPR and AUC can represent the performance of the model under different thresholds, the present invention selects AUPR and AUC as the main evaluation indicators.

[0038] In addition, the present invention also uses other threshold-based evaluation indicators, such as F1 score F1_score, accuracy Accuracy, recall rate Recall, precision Precision, and the relevant mathematical formulas are expressed as follows: ; ; ; ; Among them, TP, FP, TN and FN represent true positive, false positive, true negative and false negative, respectively.

[0039] Among them, AUC refers to the area under the operating characteristic curve, which quantitatively reflects the model performance measured based on the ROC curve; AUPR represents the area under the Precision-Recall (PR) curve; the larger the area under the PR curve, the better the performance of the model; the model performance results are shown in Table 1; Table 1 Five-fold cross validation results

[0040] In order to evaluate the prediction performance of the model, the method TransSAGE-KAN-LDA of the present invention is compared with several existing methods; the experimental results are shown in Table 2;

[0041] Table 2 Performance comparison of different methods under five-fold cross validation

[0042] Analyzing the above-mentioned validity verification results of the embodiments of the present invention, it can be seen that the lncRNA-disease association prediction method TransSAGE-KAN-LDA based on multi-layer Transformer, GraphSAGE and KAN network of the present invention is used to identify potential lncRNA-disease associations. This method fully mines five similarity information from lncRNA and disease; the present invention introduces two feature learning extractors, Transformer and GraphSAGE, from different perspectives, which are used to capture the global feature information and network topology information of lncRNA and disease respectively. Through the interaction and complementation between the modules, TransSAGE-KAN-LDA can more accurately capture the complex relationship between lncRNA and disease; under the five-fold cross validation on the data set, TransSAGE-KAN-LDA of the present invention achieved AUC values ​​of 0.9868 and 0.9803, respectively, surpassing other models.

[0043] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and the embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily realized. Therefore, without departing from the general concept defined by the claims and equivalent scope, the present invention is not limited to the specific details and the illustrations shown and described here.

Claims

1. A lncRNA-disease association prediction method based on KAN network, characterized in that: The following steps are involved: S1, integrating lncRNA-disease association network information and multiple similarity information of lncRNA and disease to construct a comprehensive heterogeneous network; S2. Construct a Transformer feature extractor based on a multi-head attention mechanism to capture global serialization features from lncRNA and disease nodes; S3. Build a GraphSAGE feature extractor based on heterogeneous graph networks to capture the topological structural features of lncRNA and disease nodes. S4. The global serialization features and topological structure features are integrated to obtain the comprehensive lncRNA-disease association pair feature information. A KAN network is constructed to predict the association score between lncRNA and disease. The KAN network includes a B-Spline transformation layer and a deep KAN structure. The learnable B-spline interpolation of the transformation layer is used to learn the nonlinear mapping relationship of the lncRNA-disease association pair feature information. The deep KAN structure is used to predict the association score between lncRNA and disease.

2. The method for predicting lncRNA-disease association based on KAN network according to claim 1, characterized in that: In step S1, an m×n association network is used to represent the association information of lncRNA-disease, where m represents the number of lncRNAs and n represents the number of diseases; When lncRNA i and disease j When there is a correlation, the corresponding position of A (l i , d j ) is 1, otherwise it is 0; the association network definition is as follows: ； Calculate the preliminary characteristic information of lncRNA and disease, including lncRNA sequence similarity lS seq , lncRNA functional similarity lS func and lncRNA Gaussian interaction profile kernel similarity lS GIPK , Disease semantic similarity DS sem and disease Gaussian interaction spectrum kernel similarity DS GIPK ; The preliminary feature information of lncRNA and disease were fused respectively to obtain the lncRNA comprehensive similarity feature LS and the disease comprehensive similarity feature DS; among them, , ; By integrating the lncRNA-disease association network information, the comprehensive similarity feature lS and the disease comprehensive similarity feature DS, a lncRNA-disease heterogeneous network was constructed: , where lS and DS represent the comprehensive similarity feature matrix of lncRNA and the comprehensive similarity feature matrix of disease, respectively, and A represents the adjacency relationship matrix between lS and DS. T Represents the transpose of the adjacency matrix.

3. The lncRNA-disease association prediction method based on KAN network according to claim 2, characterized in that: lncRNA sequence similarity seq In the calculation process, the Smith-Waterman sequence alignment algorithm was used to evaluate the sequence similarity between lncRNAs; The formula for calculating sequence similarity is as follows: ； Among them, l i and l j denote the i-th lncRNA and the j-th lncRNA, respectively, SW(l i , l j ) represents lncRNA l calculated based on Smith-Waterman alignment algorithm i and l j The sequence alignment score between .

4. The lncRNA-disease association prediction method based on KAN network according to claim 3, characterized in that: lncRNA functional similarity lS func In the calculation process, the formula is expressed as: ； Among them, p and q represent the number of diseases associated with lncRNA l1 and l2, respectively, de represents a disease, DE represents a group of diseases, DE={de1, de2, ...}, and S(de, DE) is used to calculate the relationship between diseases, which is expressed as: .

5. The lncRNA-disease association prediction method based on KAN network according to claim 4, characterized in that: lncRNA Gaussian interaction profile kernel similarity lS GIPK In the calculation process, the GIPK nuclear similarity calculation formula between lncRNAs is as follows: ； Among them, A(l i ,) and A(l j ,) represent the i-th row vector and j-th row vector of A respectively; λ n is the kernel width coefficient, defined as: , where N represents the total number of lncRNAs, A(l k ,) represents the k-th row vector of A.

6. The lncRNA-disease association prediction method based on KAN network according to claim 5, characterized in that: Disease semantic similarity DS sem In the calculation process, the disease ontology DO is used to measure the semantic similarity between related diseases, which is represented as a directed acyclic graph. By utilizing the hierarchical structure of diseases in DO, the semantic similarity is calculated based on the directed acyclic graph. The calculation formula is expressed as: ； Among them, d i and d j represent the i-th disease and the j-th disease respectively, T i and T j denotes a set of sets containing the i-th and j-th diseases, respectively, S di (t) represents the semantic impact of disease t on the i-th disease.

7. The method for predicting lncRNA-disease association based on KAN network according to claim 6, characterized in that: Disease Gaussian Interaction Spectrum Kernel Similarity DS GIPK In the calculation process, the calculation formula is expressed as: ； Among them, A(,d i ) and A(,d j ) represent the i-th column vector and j-th column vector of A respectively, λ d It is expressed as: , D represents the total number of diseases, A(,n l ) represents the lth column vector of A.

8. The lncRNA-disease association prediction method based on KAN network according to claim 7, characterized in that: In step S2, the Transformer feature extractor includes three encoding layers. In the hth attention head of one of the encoding layers, three linear transformation matrices are included: query matrix 、key matrix and value matrix , the specific linear transformation is: ； in, Represents the weight matrix of the query matrix; Represents the weight matrix of the key matrix; Represents the weight matrix of the value matrix; H l-1 Represents the output of the l-1th layer; Calculate the attention score matrix , the output of each layer is obtained by normalizing the attention score matrix and the value matrix. The specific formula is expressed as: ； Where d represents and The matrix dimensions of represents the output of the lth layer in the hth attention head; The final result of each layer is obtained by concatenating the results of each attention head through multiple attention heads; after passing through the output of the three-layer Transformer encoding layer, the feature representation F1 is extracted.

9. The lncRNA-disease association prediction method based on KAN network according to claim 8, characterized in that: In step S3, the GraphSAGE feature extractor consists of three layers of GraphSAGE calculation layers, and feature extraction is achieved through neighbor sampling and aggregation mechanism; for each lncRNA and disease node, its neighbor node set is randomly sampled, and the mean of its features is calculated, which is expressed as: ； Among them, W k is the weight matrix, u is the set of adjacent nodes of node v represented by N(v), MEAN represents the average operation, σ is the nonlinear activation function, and RELU is used; It is the node representation after calculation of the kth layer; By adopting multi-layer calculations, the output of each layer is used as the input of the next layer, and finally the representation feature matrix F2 of the heterogeneous network is obtained.

10. The lncRNA-disease association prediction method based on KAN network according to claim 9, characterized in that: In the step of fusing the global sequence features and the topological structure features to obtain the comprehensive lncRNA-disease association pair feature information, the fusion method is expressed as follows: ； Among them, coefficients α1 and α2 represent the importance of features; The calculation method of the B-Spline transformation layer is expressed as: y * =W·f(x)+b; Among them, f(x) represents the feature after B-Spline interpolation, W and b are trainable weights and bias terms; The loss function is expressed as follows: ； Among them, y i represents the true label of the lncRNA-disease association pair, represents the predicted label of the lncRNA-disease association pair, and N represents the number of lncRNA-disease association pairs.

Images