A Method and System for Predicting Maize Genome-Phenomenon Associations Based on Heterogeneous Genetic Networks

Abstract

This disclosure provides a method and system for predicting gene-phenotype associations in maize based on heterogeneous genetic networks, belonging to the field of bioinformatics. The scheme processes multi-omics data of maize to obtain a heterogeneous genetic network of attributes for multiple sets of biological molecules. Through deep matrix factorization, the network topology and node attributes are collaboratively decomposed to obtain nonlinear feature representations of multiple sets of biological molecules. In this way, network structure information and molecular attribute information are jointly mined, achieving deep fusion of multi-omics data and exploring molecular characterization and the complex interactions between molecules and phenotypes. Based on the nonlinear representation of molecules, the heterogeneous genetic network of attributes is reconstructed, and useful information in the reconstructed network is mined to achieve accurate prediction of gene-phenotype associations and complementary enhancement of association mapping across multi-omics data.

Description

Technical Field

[0001] This disclosure belongs to the field of bioinformatics technology, and in particular relates to a method and system for predicting maize gene-phenotype associations based on heterogeneous genetic networks. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Genes are the primary genetic material of organisms, supporting the basic structure and function of life; phenotypes are the morphological, structural, physiological, and biochemical characteristics exhibited by individuals with specific genotypes under certain environmental conditions. Gene-phenotype association prediction can reveal the potential genetic links between genes and phenotypes, guiding the analysis of complex traits in organisms. It is of great significance to fields such as pathological research, drug development, and crop breeding, and is a key task in biology. Therefore, how to accurately predict gene-phenotype associations using machine learning techniques is a current challenge for innovation.

[0004] Due to the many-to-many relationship between genes and phenotypes, gene-phenotype association prediction is a typical multi-label classification problem. Currently, many machine learning models based on multi-label classification frameworks are used to predict gene-phenotype associations. Since biological metabolism is a complex process, biological phenotypes are influenced by multiple genetic factors at the gene, transcription, protein, and metabolic levels. Comprehensive utilization of omics data from these different levels can help improve the effectiveness of gene-phenotype association prediction. However, existing gene-phenotype association prediction methods have certain limitations. These methods typically only utilize genomic and phenomic data, lacking effective utilization of other omics data, thus failing to achieve ideal prediction results. Secondly, these methods are generally based on network topology structures, and information loss occurs when attribute data is converted into network data. These methods fail to fully utilize attribute data, thus limiting the comprehensiveness of the prediction results. Summary of the Invention

[0005] To address the aforementioned problems, this disclosure provides a method and system for predicting maize gene-phenotype associations based on heterogeneous genetic networks. The scheme processes multi-omics data of maize to obtain heterogeneous genetic networks representing the attributes of multiple sets of biomolecules. Through deep matrix factorization, the network topology and node attributes are collaboratively decomposed to obtain nonlinear feature representations of multiple sets of biomolecules. In this way, network structure information and molecular attribute information are jointly mined, achieving deep fusion of multi-omics data and exploring molecular characterization and the complex interactions between molecules and phenotypes. Based on the nonlinear representation of molecules, the heterogeneous genetic network is reconstructed, and useful information within the reconstructed network is mined to achieve gene-phenotype association prediction.

[0006] According to a first aspect of the embodiments of this disclosure, a method for predicting maize gene-phenotype associations based on heterogeneous genetic networks is provided, comprising:

[0007] Acquire gene sequence data, gene ontology annotation data, known phenotypic tag annotation data, gene-miRNA association matrix, gene-pathway association matrix, miRNA sequence data, phenotypic tag annotation data, and phenotypic tag hierarchical structure data;

[0008] Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data; gene ontology annotation data is dimensionality reduced as gene functional attribute data; hierarchical structure attribute information of phenotypic tags is obtained based on the hierarchical structure of phenotypic tags; and miRNA sequence similarity network is obtained through miRNA sequence data.

[0009] A molecular property heterogeneous genetic network is constructed. By deeply decomposing the molecular property heterogeneous genetic network and synergistically fusing the structural data of the molecular property heterogeneous genetic network and the property data of the molecules, multiple sets of nonlinear feature representations of biological molecules are obtained.

[0010] Based on the nonlinear characteristic representation of multiple sets of biological molecules, a heterogeneous genetic network of molecular attributes is reconstructed to obtain a gene-phenotype association prediction matrix that integrates multi-omics data.

[0011] According to a second aspect of the present disclosure, a maize gene-phenotype association prediction system based on heterogeneous genetic networks is provided, comprising:

[0012] The data acquisition module is configured to acquire gene sequence data, gene ontology annotation data, known phenotypic tag annotation data, gene-miRNA association matrix, gene-pathway association matrix, miRNA sequence data, phenotypic tag annotation data, and phenotypic tag hierarchical structure data.

[0013] The network construction and feature learning module is configured to: encode gene sequence data into equal-length encoded data as gene sequence attribute data; reduce the dimensionality of gene ontology annotation data as gene functional attribute data; obtain hierarchical structure attribute information of phenotypic labels based on the hierarchical structure of phenotypic labels; obtain miRNA sequence similarity network through miRNA sequence data; construct a molecular attribute heterogeneous genetic network, and through deep decomposition of the molecular attribute heterogeneous genetic network, synergistically fuse the structural data of the molecular attribute heterogeneous genetic network and the attribute data of molecules to obtain nonlinear feature representations of multiple sets of biological molecules;

[0014] The association prediction module is configured to: reconstruct a heterogeneous genetic network of molecular attributes based on the nonlinear feature representations of multiple sets of biological molecules, and obtain a gene-phenotype association prediction matrix that integrates multi-omics data.

[0015] According to a third aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the memory. When the processor executes the program, it implements a maize gene-phenotype association prediction method based on heterogeneous genetic networks as described above.

[0016] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements a maize gene-phenotype association prediction method based on heterogeneous genetic networks as described above.

[0017] Compared with the prior art, the beneficial effects of this disclosure are:

[0018] This disclosure provides a method and system for predicting maize gene-phenotype associations based on heterogeneous genetic networks. The scheme processes multi-omics data to construct a heterogeneous genetic network of attributes for multiple sets of biological molecules. By fusing molecular network structure information with molecular attribute information and combining it with deep matrix factorization technology, nonlinear feature representations of multiple sets of biological molecules are obtained, enabling in-depth mining of molecular nonlinear features. Based on the feature representations of molecules, the heterogeneous genetic network of attributes is reconstructed to achieve gene-phenotype association prediction, improving the efficiency and accuracy of association prediction.

[0019] Advantages of this disclosure in additional aspects will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0020] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0021] Figure 1 This is a flowchart of a maize gene-phenotype association prediction method based on heterogeneous genetic networks as described in the embodiments of this disclosure;

[0022] Figure 2 This is a schematic diagram of the structure of a maize gene-phenotype association prediction system based on heterogeneous genetic networks as described in the embodiments of this disclosure. Detailed Implementation

[0023] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0026] Where there is no conflict, the embodiments and features described herein can be combined with each other.

[0027] Example 1:

[0028] The purpose of this embodiment is to provide a method for predicting maize gene-phenotype associations based on heterogeneous genetic networks.

[0029] A method for predicting maize gene-phenotype associations based on heterogeneous genetic networks includes:

[0030] Acquire gene sequence data, gene ontology annotation data, known phenotypic tag annotation data, gene-miRNA association matrix, gene-pathway association matrix, miRNA sequence data, phenotypic tag annotation data, and phenotypic tag hierarchical structure data;

[0031] Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data; gene ontology annotation data is dimensionality reduced as gene functional attribute data; hierarchical structure attribute information of phenotypic tags is obtained based on the hierarchical structure of phenotypic tags; and miRNA sequence similarity network is obtained through miRNA sequence data.

[0032] A molecular property heterogeneous genetic network is constructed. By deeply decomposing the molecular property heterogeneous genetic network and synergistically fusing the structural data of the molecular property heterogeneous genetic network and the property data of the molecules, multiple sets of nonlinear feature representations of biological molecules are obtained.

[0033] Based on the nonlinear characteristic representation of multiple sets of biological molecules, a heterogeneous genetic network of molecular attributes is reconstructed to obtain a gene-phenotype association prediction matrix that integrates multi-omics data.

[0034] Furthermore, the construction of the molecular property heterogeneous genetic network specifically involves:

[0035] Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data;

[0036] Dimensionality reduction of gene ontology annotation data is used as gene functional attribute data;

[0037] The hierarchical structure attribute information of the phenotypic tags is obtained through the hierarchical structure of the phenotypic tags;

[0038] A miRNA sequence similarity network was obtained using miRNA sequence data.

[0039] Furthermore, by performing deep decomposition on the heterogeneous genetic network of molecular attributes and synergistically fusing the structural data of the heterogeneous genetic network of molecular attributes and the attribute data of the molecules, multiple sets of nonlinear feature representations of biological molecules are obtained, specifically as follows:

[0040] Initialize low-dimensional feature representations of multiple sets of biological molecules;

[0041] For each group of biomolecules, a multi-layer neural network is designed, and the low-dimensional nonlinear feature representation and property harmonic matrix of the biomolecules are learned based on the multi-layer neural network.

[0042] Based on deep matrix factorization, we integrate molecular attribute heterogeneous genetic network data and design the computational loss of the model.

[0043] Furthermore, the step of reconstructing a heterogeneous genetic network of molecular attributes based on the nonlinear feature representations of multiple sets of biological molecules to obtain a gene-phenotype association prediction matrix that integrates multi-omics data specifically involves: reconstructing a gene-phenotype association network based on the low-dimensional nonlinear feature representations of genes and low-dimensional nonlinear feature representations of phenotypes to obtain a gene-phenotype association prediction matrix that integrates multi-omics data.

[0044] Specifically, for ease of understanding, the solution described in this embodiment will be explained in detail below with reference to the accompanying drawings:

[0045] To address the problems existing in the prior art, this embodiment provides a method for predicting maize gene-phenotype associations based on heterogeneous genetic networks, including:

[0046] S101: Obtain gene sequence data, gene ontology annotation data, and known phenotypic tag annotation data; obtain the association matrix between genes and miRNAs; obtain the association matrix between genes and pathways; obtain miRNA sequence data and phenotypic tag annotation data; obtain the hierarchical structure data of phenotypic tags.

[0047] Specifically, step 101 is implemented as follows:

[0048] S1011, the acquisition of gene sequence data, gene ontology annotation data, and known phenotypic tag annotation data;

[0049] Sequence data refers to the nucleotide sequence of a gene.

[0050] Among them, gene ontology annotation data refers to the annotation status of genes by gene ontology. Each gene has a gene ontology annotation vector, the length of which is the number of all gene ontology. When a gene is annotated by the i-th gene ontology, the i-th element of its annotation vector is 1, and all ancestor ontology i in the hierarchical directed acyclic graph are also 1 at the position corresponding to the gene annotation vector. If a gene is not annotated by the i-th gene ontology, the i-th element of its annotation vector is 0.

[0051] The known phenotypic label annotation data refers to the annotation status of genes for phenotypic labels. Each gene has a phenotypic annotation vector, the length of which is the number of all phenotypic labels. When a gene has the i-th phenotype, the i-th element of its annotation vector is 1, and all ancestral labels of phenotypic label i in the hierarchical directed acyclic graph are also 1 at the corresponding positions of the gene's annotation vector; if it does not have the i-th phenotype, the i-th element of its annotation vector is 0.

[0052] S1012, obtaining the association matrix between genes and miRNAs;

[0053] The gene-miRNA association matrix refers to the association matrix formed based on the biological phenomenon of miRNA regulating genes. The number of rows in the matrix is ​​the number of all genes, and the number of columns in the matrix is ​​the number of all miRNAs. When the i-th gene and the j-th miRNA are associated, the element in the i-th row and j-th column of the association matrix is ​​1, otherwise it is 0.

[0054] S1013, obtaining the association matrix between genes and pathways;

[0055] The gene-pathway association matrix is ​​a matrix formed based on the relationship between genes and pathways. The number of rows in the matrix is ​​the total number of genes, and the number of columns is the total number of pathways. When the i-th gene and the j-th pathway are associated, the element in the i-th row and j-th column of the association matrix is ​​1; otherwise, it is 0.

[0056] S1014, the acquisition of miRNA sequence data and phenotypic tag annotation data;

[0057] Here, sequence data refers to the nucleotide sequence of miRNA;

[0058] Phenotypic tag annotation data refers to the annotation status of miRNAs for phenotypic tags. Each miRNA has a phenotypic annotation vector, the length of which is the number of all phenotypic tags. When a miRNA has the i-th phenotype, the i-th element of its annotation vector is 1, and all ancestor tags of phenotypic tag i in the hierarchical directed acyclic graph are also 1 at the corresponding positions of the miRNA's annotation vector; if it does not have the i-th phenotype, the i-th element of its annotation vector is 0.

[0059] S1015, the hierarchical structure data of the phenotypic labels is obtained;

[0060] Among them, the phenotypic labels have a hierarchical structure and are organized in a structured way through a directed acyclic graph;

[0061] S102: Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data; gene ontology annotation data is reduced in dimensionality as gene functional attribute data; hierarchical structure attribute information of phenotypic tags is obtained through the hierarchical structure of phenotypic tags; miRNA sequence similarity network is obtained through miRNA sequence data; molecular attribute heterogeneous genetic network is constructed; by performing deep decomposition of the heterogeneous network and synergistically fusing the structural data of the heterogeneous network and the attribute data of molecules, nonlinear feature representations of multiple sets of biological molecules are obtained.

[0062] Specifically, step 102 is implemented as follows:

[0063] S1021, the gene sequence data is encoded into equal-length coding data as gene sequence attribute data. Specifically, the nucleotide sequence of the gene is encoded using the k-mer method to obtain 256-dimensional coding data for the nucleotide sequence of each gene.

[0064] S1022, dimensionality reduction of gene ontology annotation data to obtain gene functional attribute data. Specifically, gene ontology annotation data is dimensionality reduced using PCA (principal component analysis, with a specified dimension of 256) to obtain gene functional attribute data.

[0065] S1023, Obtain the hierarchical structure attribute information of the phenotypic tags through the hierarchical structure of the phenotypic tags. Specifically, obtain the hierarchical structure attribute information of the phenotypic tags by implementing the DeepWalk algorithm on the hierarchical structure of the phenotypic tags.

[0066] S1024, Obtain the miRNA sequence similarity network from miRNA sequence data. Specifically, use the "pairwiseAlignment" method in the R package Biostring to calculate the similarity between any two miRNA nucleotide sequences, and normalize it to obtain the miRNA sequence similarity network.

[0067] S1025, Construct a heterogeneous genetic network with molecular attributes. Specifically, based on the network structure data and molecular attribute data obtained and processed in steps 101 and 1021-1024, a multi-omics heterogeneous genetic network with molecular attributes is constructed. The association matrix between genes and miRNAs is obtained using R... 12 This indicates that the gene-pathway association matrix uses R... 13 This indicates that the gene phenotypic tag annotation data uses R. 14 This indicates that the miRNA sequence similarity network uses R... 22 This indicates that the miRNA phenotypic tag annotation data was generated using R. 24 This indicates that the hierarchical structure of phenotypic labels uses R. 44 This indicates that gene sequence attribute data uses X. 11 This indicates that gene functional attribute data uses X. 12 This indicates that the hierarchical attribute information of the phenotypic label uses X. 41 In summary, we obtain the set of correlation matrices. Attribute matrix set

[0068] S1026, by performing deep decomposition of heterogeneous networks and synergistically fusing structural data of heterogeneous networks and molecular property data, obtains multiple sets of nonlinear characteristic representations of biological molecules.

[0069] First, based on molecular attribute heterogeneous genetic networks, the PCA algorithm is used to initialize the low-dimensional feature representation of genes. Low-dimensional feature representation of miRNA Low-dimensional feature representation of pathways Low-dimensional feature representation of phenotypic labels based on Initialize the harmonic matrix of gene sequence properties Harmonization matrix of gene functional properties Harmonic matrix of hierarchical attributes of phenotypic tags

[0070] Secondly, for each group of biomolecules, a multi-layer neural network is designed to learn the low-dimensional nonlinear feature representation and property harmonic matrix of the biomolecules. Taking the i-th biomolecule as an example:

[0071] The calculation formula for the low-dimensional nonlinear characteristic representation of the i-th biomolecule is as follows:

[0072]

[0073] Among them, W i1 and b i1W is the weight matrix and bias of the i-th biomolecule in the first layer of the neural network. i2 and b i2 It is the weight matrix and bias of the i-th biomolecule in the second layer of the neural network, and so on. Let θ represent the activation function of the l-th layer network of the i-th biomolecule, and let θ represent the parameters of the model.

[0074] The formula for calculating the harmonic matrix of the t-th attribute of the i-th biomolecule is as follows:

[0075]

[0076] Among them, W it1 and b it1 W is the weight matrix and bias of the harmonic matrix of the t-th attribute of the i-th biomolecule in the first layer of the neural network. it2 and b it2 It is the weight matrix and bias of the harmonic matrix of the t-th attribute of the i-th biomolecule in the second layer of the neural network, and so on. Let θ represent the activation function of the l-th layer network of the harmonic matrix of the t-th attribute of the i-th biomolecule, and let θ represent the parameters of the model.

[0077] In this way, the low-dimensional nonlinear characteristic representation and property harmonic matrix of biomolecules were learned.

[0078] Finally, based on deep matrix factorization, we integrate molecular attribute heterogeneous genetic network data and design the computational loss of the model.

[0079] For network data, the correlation matrix of biomolecules i and j is reconstructed using the low-dimensional nonlinear feature representations of biomolecule i and j, and the loss is calculated using the following formula:

[0080]

[0081] Where ⊙ represents the Hadama product, J ij It is R ij The indicator matrix and its dimensions are the same as R. ij The same, if R ij (s,t)>0, then H ij (s,t)=1, otherwise, H ij (s,t)=0. G i and G j These represent the low-dimensional nonlinear characteristics of the i-th and j-th biomolecules, respectively. It is a constraint on the association of unknown noise.

[0082] For the attribute data, the t-th attribute data of the i-th biomolecule is reconstructed using the low-dimensional nonlinear feature representation of the i-th biomolecule and its t-th attribute harmonic matrix, and the loss is calculated using the following formula:

[0083]

[0084] Among them, U it The harmonic matrix represents the t-th property of the i-th biomolecule.

[0085] Combining the two losses mentioned above, we obtain the total loss of the objective function:

[0086]

[0087] Here, θ stores the parameters of the neural network. Through this method, the network data and molecular attribute data in the heterogeneous genetic network of molecular attributes are co-decomposed, enabling deep fusion of multi-omics data and obtaining low-dimensional nonlinear feature representations G of multiple omics biomolecules. i .

[0088] S103. Based on the nonlinear characteristic representation of multiple sets of biological molecules, the attribute heterogeneous genetic network is reconstructed to obtain the gene-phenotype association prediction matrix that integrates multi-omics data.

[0089] Specifically, step 103 is implemented as follows:

[0090] Based on the gene low-dimensional nonlinear feature representation G1 and the phenotype low-dimensional nonlinear feature representation G4 obtained in step 102, the gene-phenotype association network is reconstructed. Obtain the gene-phenotype association prediction matrix by fusing multi-omics data. Example 2:

[0091] The purpose of this embodiment is to provide a maize gene-phenotype association prediction system based on heterogeneous genetic networks.

[0092] A maize gene-phenotype association prediction system based on heterogeneous genetic networks includes:

[0093] The data acquisition module is configured to acquire gene sequence data, gene ontology annotation data, known phenotypic tag annotation data, gene-miRNA association matrix, gene-pathway association matrix, miRNA sequence data, phenotypic tag annotation data, and phenotypic tag hierarchical structure data.

[0094] The network construction and feature learning module is configured to: encode gene sequence data into equal-length encoded data as gene sequence attribute data; reduce the dimensionality of gene ontology annotation data as gene functional attribute data; obtain hierarchical structure attribute information of phenotypic labels based on the hierarchical structure of phenotypic labels; obtain miRNA sequence similarity network through miRNA sequence data; construct a molecular attribute heterogeneous genetic network, and through deep decomposition of the molecular attribute heterogeneous genetic network, synergistically fuse the structural data of the molecular attribute heterogeneous genetic network and the attribute data of molecules to obtain nonlinear feature representations of multiple sets of biological molecules;

[0095] The association prediction module is configured to: reconstruct a heterogeneous genetic network of molecular attributes based on the nonlinear feature representations of multiple sets of biological molecules, and obtain a gene-phenotype association prediction matrix that integrates multi-omics data.

[0096] Furthermore, the system described in this embodiment corresponds to the method described in Embodiment 1, and its technical details have been described in detail in Embodiment 1, so they will not be repeated here.

[0097] In further embodiments, the following is also provided:

[0098] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When executed by the processor, the computer instructions perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0099] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0100] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0101] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0102] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0103] Those skilled in the art will recognize that the units, i.e., algorithm steps, of the various examples described in connection with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art 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 disclosure.

[0104] The above embodiments provide a maize gene-phenotype association prediction method and system based on heterogeneous genetic networks, which can be implemented and has broad application prospects.

[0105] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for predicting maize gene-phenotype associations based on heterogeneous genetic networks, characterized in that, include: Acquire gene sequence data, gene ontology annotation data, known phenotypic tag annotation data, gene-miRNA association matrix, gene-pathway association matrix, miRNA sequence data, phenotypic tag annotation data, and phenotypic tag hierarchical structure data; Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data; gene ontology annotation data is dimensionality reduced as gene functional attribute data; hierarchical structure attribute information of phenotypic tags is obtained based on the hierarchical structure of phenotypic tags; and miRNA sequence similarity network is obtained through miRNA sequence data. A molecular property heterogeneous genetic network is constructed. By deeply decomposing the molecular property heterogeneous genetic network and synergistically fusing the structural data of the molecular property heterogeneous genetic network and the property data of the molecules, multiple sets of nonlinear feature representations of biological molecules are obtained. Based on the nonlinear characteristic representation of multiple sets of biological molecules, a heterogeneous genetic network of molecular attributes is reconstructed to obtain a gene-phenotype association prediction matrix that integrates multi-omics data. The construction of the molecular property heterogeneous genetic network specifically involves: Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data; Dimensionality reduction of gene ontology annotation data is used as gene functional attribute data; The hierarchical structure attribute information of the phenotypic tags is obtained through the hierarchical structure of the phenotypic tags; A miRNA sequence similarity network was obtained using miRNA sequence data; The process involves deep decomposition of the molecular property heterogeneous genetic network, synergistic fusion of the structural data of the molecular property heterogeneous genetic network and the property data of the molecules, to obtain multiple sets of nonlinear feature representations of biological molecules, specifically: Initialize low-dimensional feature representations of multiple sets of biological molecules; For each group of biomolecules, a multi-layer neural network is designed, and the low-dimensional nonlinear feature representation and property harmonic matrix of the biomolecules are learned based on the multi-layer neural network. Based on deep matrix factorization, we integrate molecular attribute heterogeneous genetic network data and design the computational loss of the model.

2. The method for predicting maize gene-phenotype associations based on heterogeneous genetic networks as described in claim 1, characterized in that, The process involves reconstructing a heterogeneous genetic network of molecular attributes based on the nonlinear feature representations of multiple sets of biological molecules, and obtaining a gene-phenotype association prediction matrix that integrates multi-omics data. Specifically, this involves reconstructing a gene-phenotype association network based on the low-dimensional nonlinear feature representations of genes and low-dimensional nonlinear feature representations of phenotypes, and obtaining a gene-phenotype association prediction matrix that integrates multi-omics data.

3. A maize gene-phenotype association prediction system based on heterogeneous genetic networks, characterized in that, include: The data acquisition module is configured to acquire gene sequence data, gene ontology annotation data, known phenotypic tag annotation data, gene-miRNA association matrix, gene-pathway association matrix, miRNA sequence data, phenotypic tag annotation data, and phenotypic tag hierarchical structure data. The network construction and feature learning module is configured to: encode gene sequence data into equal-length encoded data as gene sequence attribute data; reduce the dimensionality of gene ontology annotation data as gene functional attribute data; obtain hierarchical structure attribute information of phenotypic labels based on the hierarchical structure of phenotypic labels; obtain miRNA sequence similarity network through miRNA sequence data; construct a molecular attribute heterogeneous genetic network, and through deep decomposition of the molecular attribute heterogeneous genetic network, synergistically fuse the structural data of the molecular attribute heterogeneous genetic network and the attribute data of molecules to obtain nonlinear feature representations of multiple sets of biological molecules; The association prediction module is configured to: reconstruct a heterogeneous genetic network of molecular attributes based on the nonlinear feature representations of multiple sets of biological molecules, and obtain a gene-phenotype association prediction matrix that integrates multi-omics data; The construction of the molecular property heterogeneous genetic network specifically involves: Gene sequence data is encoded into equal-length encoded data as gene sequence attribute data; Dimensionality reduction of gene ontology annotation data is used as gene functional attribute data; The hierarchical structure attribute information of the phenotypic tags is obtained through the hierarchical structure of the phenotypic tags; A miRNA sequence similarity network was obtained using miRNA sequence data; The process involves deep decomposition of the molecular property heterogeneous genetic network, synergistic fusion of the structural data of the molecular property heterogeneous genetic network and the property data of the molecules, to obtain multiple sets of nonlinear feature representations of biological molecules, specifically: Initialize low-dimensional feature representations of multiple sets of biological molecules; For each group of biomolecules, a multi-layer neural network is designed, and the low-dimensional nonlinear feature representation and property harmonic matrix of the biomolecules are learned based on the multi-layer neural network. Based on deep matrix factorization, we integrate molecular attribute heterogeneous genetic network data and design the computational loss of the model.

4. The maize gene-phenotype association prediction system based on heterogeneous genetic networks as described in claim 3, characterized in that, The process involves reconstructing a heterogeneous genetic network of molecular attributes based on the nonlinear feature representations of multiple sets of biological molecules, and obtaining a gene-phenotype association prediction matrix that integrates multi-omics data. Specifically, this involves reconstructing a gene-phenotype association network based on the low-dimensional nonlinear feature representations of genes and low-dimensional nonlinear feature representations of phenotypes, and obtaining a gene-phenotype association prediction matrix that integrates multi-omics data.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running thereon, characterized in that, When the processor executes the program, it implements the maize gene-phenotype association prediction method based on heterogeneous genetic networks as described in any one of claims 1-2.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the maize gene-phenotype association prediction method based on heterogeneous genetic networks as described in any one of claims 1-2.

Images