Apparatus and method for diagnosing genetic risk of alzheimer's disease on basis of SNP interaction data

Abstract

An apparatus for diagnosing the genetic risk of Alzheimer's disease according to the present disclosure comprises: a data acquisition unit for acquiring single nucleotide polymorphism (SNP) data from a blood sample; a data processing unit for acquiring independent data of each of SNPs on the basis of the SNP data, acquiring first interaction data between the SNPs from the SNP data by means of statistical analysis, and acquiring second interaction data between the SNPs from a gene network composed of respective genes corresponding to the SNPs; a neural network processing unit for generating a graph-based neural network model reflecting the independent data, the first interaction data, and the second interaction data, and training the graph-based neural network model by adjusting parameters of the graph-based neural network model; and a diagnosis unit for diagnosing the genetic risk of Alzheimer's disease by using the trained graph-based neural network model.

Description

Device and method for diagnosing genetic risk of Alzheimer's disease based on SNP interaction data

[0001] The present invention relates to an apparatus and method for diagnosing genetic risk of Alzheimer's disease based on SNP interaction data. More specifically, the invention relates to an apparatus and method for diagnosing genetic risk of Alzheimer's disease using a graph-based neural network model that reflects independent data of SNPs and interaction data between SNPs.

[0002] The present invention is derived from research conducted as part of the Ministry of Science and ICT’s Individual Basic Research (MSIT) (Project Unique Number: 2710085913, Project Number: 2021R1A2C2003474, Research Project Name: Development of an Artificial Intelligence Model for Precision Medicine and Diagnosis Diversification of Dementia, Project Performing Institution Name: Ajou University).

[0003] The present invention was derived from research conducted as part of the Ministry of Science and ICT’s Artificial Intelligence Convergence Innovation Talent Development (R&D) (Project Unique Number: 2710033934, Project Number: RS-2023-00255968, Research Project Name: Artificial Intelligence Convergence Innovation Talent Development (Ajou University), Project Performing Organization Name: Ajou University Industry-Academic Cooperation Foundation).

[0004] The present invention is derived from research conducted as part of the Ministry of Education's establishment of a research infrastructure for science and engineering (Project No.: 2340032323, Project No.: 2022R1A6A3A01086784, Research Project Title: Development of a Multimodal-Multi-Domain Based Machine Learning Algorithm for Predicting Dementia Progression, Project Performing Institution: Ajou University).

[0005] The present invention was derived from research conducted as part of the Ministry of Health and Welfare's Global Training Support for Medical Scientists (Project No.: 2460002543, Project No.: RS-2024-00407544, Research Project Title: Development of a New Anticancer Treatment Strategy Based on Liver Cancer-TME Connectome, Project Performing Organization: Ajou University Industry-Academic Cooperation Foundation).

[0006] The present invention is derived from research conducted as part of the Ministry of Science and ICT’s Individual Basic Research (MSIT) (R&D) (Project No.: 2710078560, Project No.: RS-2022-00165386, Research Project Title: Discovery of Epidrivers to Control Cancer Progression and Development of Diagnostic and Therapeutic Candidates, Project Performing Organization: Ajou University Industry-Academic Cooperation Foundation).

[0007] The present invention was derived from research conducted as part of the Ministry of Health and Welfare's project to foster convergence talents specialized in medical artificial intelligence (Project No.: 2460004446, Project No.: RS-2025-02310331, Research Project Name: Ajou University Medical Artificial Intelligence Specialized Convergence Talent Development Project Group, Project Performing Organization Name: Ajou University Industry-Academic Cooperation Foundation).

[0008] The present invention was derived from research conducted as part of the Ministry of Health and Welfare's Research-Oriented Hospital Development (R&D) (Project No.: 2460004015, Project No.: RS-2021-KH113821, Research Project Title: Establishment of Human-Environmental Interaction Beyond Target Platform, Project Performing Organization: Ajou University Industry-Academic Cooperation Foundation).

[0009]

[0010] Meanwhile, the Korean government, the provider of the problem, has no property interest in all aspects of the present invention.

[0011] Alzheimer's disease is a representative neurodegenerative disease that is emerging as a serious medical and social problem worldwide. Genetic factors significantly influence the onset and progression of Alzheimer's disease, and recent research has revealed that single nucleotide polymorphisms (SNPs) play a crucial role in assessing the polygenic risk of the disease.

[0012] Existing Polygenic Risk Score (PRS) models assessed the genetic risk associated with Alzheimer's disease in a simplified manner by summing independent SNP data. However, while these models reflect independent SNP data, they do not consider interaction data between SNPs, which limits the accuracy and reliability of their predictions.

[0013] Therefore, there is a need for a diagnostic method for the genetic risk of Alzheimer's disease that reflects not only independent SNP data but also interaction data between SNPs.

[0014] The present disclosure aims to provide an apparatus and method for diagnosing genetic risk of Alzheimer's disease with improved accuracy and reliability of prediction by simultaneously reflecting not only independent data of each SNP but also interaction data between SNPs.

[0015] In addition, the present disclosure aims to enable early treatment of Alzheimer's disease by classifying high-risk groups based on genetic risk probabilities calculated through a trained graph-based neural network model.

[0016] An Alzheimer's disease genetic risk diagnostic device according to one embodiment of the present disclosure may include: a data acquisition unit that acquires Single Nucleotide Polymorphism (SNP) data from a blood sample; a data processing unit that acquires independent data for each SNP based on the SNP data, acquires first interaction data between SNPs from the SNP data through statistical analysis, and acquires second interaction data between SNPs from a gene network composed of genes corresponding to each SNP; a neural network processing unit that generates a graph-based neural network model by reflecting the independent data, the first interaction data, and the second interaction data, and trains the graph-based neural network model by adjusting the parameters of the graph-based neural network model; and a diagnostic unit that diagnoses the genetic risk of Alzheimer's disease using the trained graph-based neural network model.

[0017] In some embodiments, the data processing unit may obtain first interaction data between SNPs by performing an epistasis test to evaluate statistical significance between SNPs based on SNP data.

[0018] In some embodiments, the data processing unit maps each SNP to a gene corresponding to each SNP and can obtain second interaction data between SNPs based on interaction data between genes in a gene network composed of genes.

[0019] In some embodiments, the neural network processing unit can generate a graph-based neural network model by defining each SNP as a node of a graph-based neural network model and defining edges of the graph-based neural network model based on first interaction data between SNPs and second interaction data between SNPs.

[0020] In some embodiments, the neural network processing unit can train a graph-based neural network model by adjusting a first parameter that determines the intensity of interaction data spread between SNPs.

[0021] In some embodiments, the neural network processing unit can train a graph-based neural network model by adjusting a second parameter that determines the combination ratio of independent data for each SNP, first interaction data between SNPs, and second interaction data between SNPs.

[0022] In some embodiments, the neural network processing unit may generate combined data in which independent data for each SNP, first interaction data between SNPs, and second interaction data between SNPs are combined according to the second parameter.

[0023] In some embodiments, the neural network processing unit can train a graph-based neural network model by adjusting a third parameter for calculating the genetic risk probability of Alzheimer's disease according to combined data.

[0024] In some embodiments, the diagnostic unit calculates the genetic risk probability of Alzheimer's disease using a trained graph-based neural network model, and can diagnose a high-risk group for Alzheimer's disease if the calculated probability value exceeds a threshold value.

[0025] A method for diagnosing genetic risk of Alzheimer's disease according to one embodiment of the present disclosure may include: a step of obtaining Single Nucleotide Polymorphism (SNP) data from a blood sample; a step of obtaining independent data for each SNP based on the SNP data; a step of obtaining first interaction data between SNPs from the SNP data through statistical analysis; a step of obtaining second interaction data between SNPs from a gene network composed of genes corresponding to each SNP; a step of generating a graph-based neural network model by reflecting the independent data, the first interaction data, and the second interaction data; a step of training a graph-based neural network model by adjusting the parameters of the graph-based neural network model; and a step of diagnosing genetic risk of Alzheimer's disease using the trained graph-based neural network model.

[0026] In some embodiments, the step of obtaining first interaction data may include the step of obtaining first interaction data between SNPs by performing an epistasis test to evaluate statistical significance between SNPs based on SNP data.

[0027] In some embodiments, the step of obtaining second interaction data may include: mapping each SNP to a gene corresponding to each SNP; and obtaining second interaction data between SNPs based on interaction data between genes in a gene network composed of genes.

[0028] In some embodiments, the step of generating a graph-based neural network model may include: defining each SNP as a node of the graph-based neural network model; and defining edges of the graph-based neural network model based on first interaction data between SNPs and second interaction data between SNPs.

[0029] In some embodiments, the step of training a graph-based neural network model may include the step of adjusting a first parameter that determines the intensity of the interaction data spread between SNPs.

[0030] In some embodiments, the step of training a graph-based neural network model may further include the step of adjusting a second parameter that determines the combination ratio of independent data for each SNP, first interaction data between SNPs, and second interaction data between SNPs.

[0031] In some embodiments, the step of training a graph-based neural network model may further include the step of generating combined data in which independent data for each SNP, first interaction data between SNPs, and second interaction data between SNPs are combined according to a second parameter.

[0032] In some embodiments, the step of training a graph-based neural network model may further include the step of adjusting a third parameter for calculating the genetic risk probability of Alzheimer's disease according to combined data.

[0033] In some embodiments, the step of diagnosing the genetic risk of Alzheimer's disease may include: a step of calculating the probability of the genetic risk of Alzheimer's disease using a learned graph-based neural network model; and a step of diagnosing a high-risk group for Alzheimer's disease if the calculated probability value exceeds a threshold value.

[0034] The device and method for diagnosing genetic risk of Alzheimer's disease according to an embodiment of the present disclosure can improve the accuracy and reliability of prediction by simultaneously reflecting not only independent data of each SNP but also interaction data between SNPs.

[0035] In addition, the device and method for diagnosing genetic risk of Alzheimer's disease according to an embodiment of the present disclosure can enable early treatment of Alzheimer's disease by classifying high-risk groups for Alzheimer's disease based on genetic risk probabilities calculated through a learned graph-based neural network model.

[0036] FIG. 1 is a drawing showing an Alzheimer's disease genetic risk diagnostic device according to an embodiment of the present disclosure.

[0037] FIG. 2 is a flowchart illustrating a method for diagnosing genetic risk of Alzheimer's disease according to an embodiment of the present disclosure.

[0038] FIG. 3 is a drawing illustrating a method for generating combined data according to an embodiment of the present disclosure.

[0039] FIG. 4 is a diagram showing a case where the intensity of interaction data diffusion between SNPs is controlled by adjusting the first parameter according to an embodiment of the present disclosure.

[0040] FIG. 5 is a diagram showing a case where the combination ratio of combination data is adjusted by adjusting a second parameter according to an embodiment of the present disclosure.

[0041] FIG. 6 is a diagram showing a case in which the genetic risk probability of Alzheimer's disease is calculated according to combined data by adjusting the third parameter according to an embodiment of the present disclosure.

[0042] FIG. 7 is a flowchart illustrating a graph-based neural network model learning method according to an embodiment of the present disclosure.

[0043] FIG. 8 is a flowchart illustrating a method for updating a graph-based neural network model according to an embodiment of the present disclosure.

[0044] Hereinafter, preferred embodiments of the present invention will be described as follows with reference to the attached drawings.

[0045] In the following, terms such as "upper," "middle," and "lower" may be replaced with other terms, such as "first," "second," and "third," to describe the components of the specification. While terms such as "first," "second," and "third" may be used to describe various components, they are not limited by these terms, and "first component" may be named "second component."

[0046]

[0047] FIG. 1 is a drawing showing an Alzheimer's disease genetic risk diagnostic device according to an embodiment of the present disclosure.

[0048] Referring to FIG. 1, an Alzheimer's disease genetic risk diagnosis device (100) according to an embodiment of the present disclosure may include a data acquisition unit (110), a data processing unit (120), a neural network processing unit (130), and a diagnosis unit (140).

[0049] The Alzheimer's disease genetic risk diagnostic device (100) may be a device that performs the acquisition, preprocessing, and analysis of SNP data, the training of a graph-based neural network model, and the derivation of a diagnostic result. The Alzheimer's disease genetic risk diagnostic device (100) can diagnose the genetic risk of Alzheimer's disease by reflecting independent data and interaction data between SNPs through a graph-based neural network model.

[0050] In the present disclosure, SNP (Single Nucleotide Polymorphism) data may refer to data representing genetic markers in which a single nucleotide is mutated in a specific DNA sequence. SNP data may represent genetic diversity among individuals and can be utilized to analyze associations with Alzheimer's disease or to assess genetic risk. Additionally, a graph-based neural network model may refer to a machine learning model that learns the interactions of data based on a graph structure composed of nodes and edges.

[0051] The data acquisition unit (110) can acquire SNP data from a blood sample. In some embodiments, the data acquisition unit (110) can isolate DNA from a blood sample and acquire SNP data from the isolated DNA. In other embodiments, the data acquisition unit (110) can acquire SNP data by utilizing a known SNP database. For example, the data acquisition unit (110) can acquire SNP data by utilizing a known database such as dbSNP, KoreanChip, etc.

[0052] Subsequently, the data acquisition unit (110) may perform preprocessing operations such as processing missing values, removing noise, and filtering unnecessary variants to select only valid SNPs. The preprocessed SNP data may undergo standardization and normalization processes to be suitable for statistical analysis and neural network learning, and the preprocessed SNP data may be transmitted to the data processing unit (120).

[0053] The data processing unit (120) can analyze SNP data to obtain independent data for each SNP and interaction data between SNPs for Alzheimer's disease.

[0054] In the present disclosure, independent data may refer to values ​​that quantitatively represent the effect of each specific SNP individually on Alzheimer's disease. Additionally, in the present disclosure, interaction data may refer to values ​​that quantitatively represent the effect of the interaction between two or more SNPs on Alzheimer's disease.

[0055] For example, the data processing unit (120) can obtain independent data for each SNP by analyzing the statistical association between the SNP and Alzheimer's disease using a Genome-Wide Association Study (GWAS) tool. Here, the Genome-Wide Association Study tool may refer to a statistical tool that analyzes large-scale genomic data to identify genetic variations associated with a specific phenotype (e.g., Alzheimer's disease). In the present disclosure

[0056] Additionally, the data processing unit (120) can obtain interaction data indicating the effect of interactions between two or more SNPs on Alzheimer's disease. For example, the interaction data may include first interaction data obtained based on statistical correlations between SNPs and second interaction data indirectly calculated based on gene-gene interaction data (GGI) after mapping the SNP data to a gene network.

[0057] The first interaction data can be obtained, for example, through an Epistasis Test. Here, the Epistasis Test is a statistical technique for evaluating the impact of an interaction between two SNPs on disease development, and may refer to a method of deriving the strength and direction of an interaction between SNPs by utilizing logistic regression analysis.

[0058] Since SNPs exist at specific locations in the genome and genes are located in specific nucleotide sequence regions within the genome, each SNP can be mapped to a specific gene, and interaction data between SNPs—in other words, secondary interaction data—can be obtained indirectly from interaction data between genes within a gene network composed of mapped genes.

[0059] The data processing unit (120) can transmit independent data for each acquired SNP, first interaction data, and second interaction data to the neural network processing unit (130).

[0060] The neural network processing unit (130) can generate and train a graph-based neural network model (131) based on data provided by the data processing unit (120).

[0061] The neural network processing unit (130) can generate a graph-based neural network model (131) by defining each SNP as a node of the graph-based neural network model (131) and defining interaction data between SNPs as edges of the graph-based neural network model (131). Here, the nodes of the graph-based neural network model (131) may include independent data for each SNP. The graph-based neural network model (131) may, for example, be a Multi-Graph Propagation Network (MGPN) that reflects two or more interaction data.

[0062] The neural network processing unit (130) can train the generated graph-based neural network model (131) by adjusting the first parameter, the second parameter, and the third parameter.

[0063] The first parameter may be a parameter that controls the intensity of the spread of interaction data between SNPs in the graph. In the present disclosure, the first parameter may be referred to as a smoothness parameter and may control the learning intensity in a direction that emphasizes interactions between nearby nodes or includes interactions between distant nodes.

[0064] For example, when the first parameter decreases, interactions between nearby nodes within the graph-based neural network model (131) may be emphasized, and when the first parameter increases, interactions between distant nodes may be considered more. The first parameter (μ ) can satisfy the following mathematical equation 1.

[0065]

[0066] However, here is the interaction data matrix, The Laplacian matrix, which mathematically represents the connection relationships between nodes in a graph, is the identity matrix, s is the number of SNPs, n is the number of samples, can be a matrix representing independent data for each SNP. Also, It may be P (Phenotypic Interaction) representing the first interaction data or G (Genomic Interaction) representing the second interaction data.

[0067] The second parameter may be a parameter that adjusts the contribution of the independent data of the SNP, the first interaction data, and the second interaction data according to a ratio. Through this, combined data can be generated by combining the independent data of the SNP, the first interaction data, and the second interaction data. In the present disclosure, combined data is data generated by combining the independent data of each SNP with the first interaction data and the second interaction data between the SNPs at a specific ratio, and can be used to predict the genetic risk of Alzheimer's disease. The second parameter (θ , I or P or G) can satisfy the following mathematical equation 2.

[0068]

[0069] step, is combined data, is independent data of SNPs is the first interaction data, is the second interaction data, s is the number of SNPs, n is the number of samples, is a second parameter that moderates independent data of SNPs is a second parameter that regulates the first interaction data is a second parameter that regulates the second interaction data.

[0070] The third parameter may be a parameter used to ultimately calculate the genetic risk probability of Alzheimer's disease by utilizing combined data. For example, the third parameter ( ) can satisfy the following mathematical equation 3.

[0071]

[0072] step, is the genetic risk probability of Alzheimer's disease, is the natural constant, is the third parameter, is combined data, is the number of samples.

[0073] The diagnostic unit (140) can calculate the genetic risk probability of Alzheimer's disease using a learned graph-based neural network model (131). The diagnostic unit (140) can calculate the genetic risk probability of Alzheimer's disease according to the above mathematical formula 3, and if the calculated probability value exceeds a preset threshold value (e.g., 0.7), the diagnostic subject can be classified as a high-risk group for Alzheimer's disease.

[0074] The Alzheimer's disease genetic risk diagnosis device (100) according to an embodiment of the present disclosure can improve the accuracy and reliability of the Alzheimer's disease genetic risk diagnosis by simultaneously reflecting not only independent data of each SNP but also interaction data between SNPs.

[0075] In addition, the Alzheimer's disease genetic risk diagnostic device (100) according to an embodiment of the present disclosure can enable early treatment of Alzheimer's disease by classifying high-risk groups for Alzheimer's disease based on genetic risk probabilities calculated through a learned graph-based neural network model.

[0076]

[0077] FIG. 2 is a flowchart illustrating a method for diagnosing genetic risk of Alzheimer's disease according to an embodiment of the present disclosure. FIG. 2 can be explained with reference to FIG. 1 described above.

[0078] Referring to FIG. 2, a method for diagnosing genetic risk of Alzheimer's disease (S100) according to an embodiment of the present disclosure may include the steps of: obtaining SNP data from a blood sample (S110); obtaining independent data for each SNP based on the SNP data (S120); obtaining first interaction data between SNPs from the SNP data through statistical analysis (S130); obtaining second interaction data between SNPs from a gene network (S140); creating a neural network model (S150); training a neural network model by adjusting parameters of the neural network model (S160); and diagnosing genetic risk of Alzheimer's disease using the trained neural network model (S170).

[0079] FIG. 2 illustrates steps S110 to S170 being performed sequentially, but is not limited thereto; some steps may be merged and performed simultaneously, some steps may be omitted, or new steps may be added.

[0080] In step (S110), SNP data can be obtained from a blood sample. In some embodiments, SNP data can be obtained using a known SNP analysis technique on the blood sample. In other embodiments, SNP data can be obtained by utilizing a known database (e.g., dbSNP, KoreanChip).

[0081] In step S120, independent data for each SNP can be obtained based on the SNP data. For example, independent data for SNPs can be obtained through a Genome-Wide Association Study (GWAS).

[0082] In step S130, first interaction data between SNPs can be obtained from SNP data through statistical analysis. For example, first interaction data can be obtained by performing an epistasis test to calculate the effect of the interaction between two SNPs on the development of Alzheimer's disease.

[0083] In step S140, second interaction data between SNPs can be obtained from a gene network composed of genes corresponding to each SNP. For example, a gene network can be generated by mapping each SNP to a gene associated with each SNP, and second interaction data between SNPs can be obtained by utilizing SNP-gene relationships and gene-gene interactions (GGI).

[0084] In step S150, a neural network model can be created by reflecting independent data, first interaction data, and second interaction data. For example, a graph-based neural network model can be created by defining each SNP as a node of a graph-based neural network model and defining first interaction data and second interaction data between SNPs as edges.

[0085] In step S160, the neural network model can be trained by adjusting the parameters of the neural network model. For example, the neural network model can be trained by adjusting a first parameter that controls the diffusion strength of interaction data, a second parameter that determines the contribution of independent data and interaction data, and a third parameter that calculates the genetic risk probability of Alzheimer's disease based on combined data.

[0086] In step S170, the genetic risk of Alzheimer's disease can be diagnosed using a trained neural network model. For example, the genetic risk probability of Alzheimer's disease can be calculated from combined data using the trained neural network model, and if the calculated probability value exceeds a preset threshold, the subject can be classified as a high-risk group for Alzheimer's disease.

[0087]

[0088] FIG. 3 is a drawing illustrating a method for generating combined data according to an embodiment of the present disclosure. FIG. 3 can be described with reference to FIG. 1 and FIG. 2 described above.

[0089] Referring to FIG. 3, combined data can be generated by combining first interaction data, second interaction data, and independent data.

[0090] The first interaction data can be obtained by performing an epistasis test. For example, a phenotypic network (121) can be generated using a GWAS tool on SNP data, and the first interaction data can be obtained by statistically analyzing the interactions between SNPs within the phenotypic network.

[0091] Second interaction data can be obtained indirectly by utilizing a gene network (122). Specifically, each SNP (S) can be mapped to a gene (G) associated with the SNP (S), and second interaction data, which is indirect interaction data between SNPs, can be obtained by utilizing a gene network (122) composed of mapped genes (G).

[0092] Independent data on the effect of each SNP on Alzheimer's disease can be obtained, for example, through a genome-wide association analysis (GWAS).

[0093] Combined data can be obtained by linearly combining independent data, first interaction data, and second interaction data.

[0094]

[0095] FIG. 4 is a diagram illustrating a case in which the intensity of interaction data diffusion between SNPs is controlled by adjusting a first parameter according to an embodiment of the present disclosure. FIG. 4 can be explained with reference to FIG. 1 to 3 described above.

[0096] A graph-based neural network model (131) can be trained by considering independent data for each SNP and interaction data between SNPs (e.g., first interaction data and second interaction data), and the training intensity can be adjusted through a first parameter to emphasize interactions between nearby nodes or to include interactions between distant nodes.

[0097] For example, if the first parameter decreases, the interaction between nearby nodes within the graph-based neural network model (131) may be emphasized, and if the first parameter increases, the interaction between distant nodes may be considered more.

[0098]

[0099] FIG. 5 is a diagram illustrating a case in which the combination ratio of combination data is adjusted by adjusting a second parameter according to an embodiment of the present disclosure. FIG. 5 can be explained with reference to FIG. 1 to FIG. 4 described above.

[0100] Referring to FIG. 5, independent data of each SNP, first interaction data, and second interaction data can be linearly combined according to a second parameter that controls the combination ratio of each data to generate combined data. The combined data generated according to the second parameter can be used to predict the genetic risk of Alzheimer's disease.

[0101]

[0102] FIG. 6 is a diagram illustrating a case in which the probability of genetic risk of Alzheimer's disease is calculated according to combined data by adjusting a third parameter according to an embodiment of the present disclosure. FIG. 6 can be explained with reference to FIG. 1 to 5 described above.

[0103] Referring to FIG. 6, the Alzheimer's disease genetic risk diagnostic device (100) can calculate the genetic risk probability of Alzheimer's disease according to a third parameter based on combined data. In some embodiments, if the calculated probability value exceeds a preset threshold, the subject to diagnosis may be classified as a high-risk group for Alzheimer's disease. On the other hand, if the calculated probability value is below the threshold, the subject to diagnosis may be classified as having mild cognitive impairment (MCI).

[0104]

[0105] FIG. 7 is a flowchart illustrating a graph-based neural network model learning method according to an embodiment of the present disclosure. FIG. 7 can be described with reference to FIG. 1 to 6 described above.

[0106] Referring to FIG. 7, a graph-based neural network model learning method (S200) according to an embodiment of the present disclosure may include a step of creating a graph-based neural network model (S210), a step of updating the graph-based neural network model by adjusting the parameters of the neural network model (S220), a step of obtaining a loss function value for evaluating the performance of the graph-based neural network model (S230), and a step of determining whether the amount of change of the loss function value is below a threshold value (S240). The graph-based neural network model learning method (S200) illustrated in FIG. 7 may, for example, correspond to steps S150 to S160 of FIG. 2.

[0107] FIG. 7 illustrates steps S210 to S240 being performed sequentially, but is not limited thereto; some steps may be merged and performed simultaneously, some steps may be omitted, or new steps may be added.

[0108] In step S210, a graph-based neural network model can be created. For example, a graph-based neural network model can be created by defining each SNP as a node of the graph-based neural network model and defining first interaction data and second interaction data between SNPs as edges.

[0109] In step S220, the graph-based neural network model can be updated by adjusting the parameters of the neural network model. For example, the first to third parameters can be set to initial values ​​and their values ​​can be adjusted by a unit value as learning is performed. As the values ​​of the first and second parameters change, the influence of the interaction data between nodes of the graph-based neural network model on the combined data may change, and as the value of the third parameter changes, the genetic risk probability value of Alzheimer's disease may change.

[0110] In step S230, a loss function value can be obtained to evaluate the performance of the graph-based neural network model. For example, known loss functions such as Cross-Entropy Loss or Mean Squared Error (MSE) may be used.

[0111] In step S240, it can be determined whether the amount of change in the loss function value is less than or equal to a threshold change amount. For example, the amount of change can be calculated by comparing the difference between the loss function value calculated in the current training step and the loss function value of the previous step, and if the amount of change in the loss function value is less than or equal to a threshold change amount (e.g., 0.001 or less), the training of the graph-based neural network model can be terminated. If the amount of change in the loss function value exceeds the threshold change amount, steps S220 through S240 can be repeated.

[0112]

[0113] FIG. 8 is a flowchart illustrating a method for updating a graph-based neural network model according to an embodiment of the present disclosure. FIG. 8 can be described with reference to FIG. 1 through 7 described above.

[0114] Referring to FIG. 8, a method (S300) for updating a graph-based neural network model according to an embodiment of the present disclosure may include a step of adjusting a first parameter (S310), a step of adjusting a second parameter (S320), a step of generating combined data (S330), and a step of adjusting a third parameter (S340).

[0115] In step S310, a first parameter determining the intensity of the interaction data between SNPs can be adjusted. For example, a smaller first parameter emphasizes interactions between nearby SNPs, while a larger first parameter may consider interactions between distant SNPs. Through this, the intensity of the diffusion related to the connection strength between nodes of a graph-based neural network model can be controlled.

[0116] In step S320, a second parameter can be adjusted to determine the combination ratio of independent data for each SNP, first interaction data between SNPs, and second interaction data between SNPs. The second parameter can determine the contribution of each data point; for example, if independent data between SNPs has a greater impact on genetic risk, the second parameter can be adjusted to increase that ratio.

[0117] In step S330, combined data can be generated by combining independent data of each SNP, first interaction data between SNPs, and second interaction data between SNPs according to the second parameter. Combined data can be generated by combining independent data of each SNP, first interaction data, and second interaction data according to the ratio determined by the second parameter.

[0118] In step S340, a third parameter for calculating the genetic risk probability of Alzheimer's disease can be adjusted according to the combined data. Based on the combined data, the genetic risk probability of Alzheimer's disease can be calculated according to the third parameter.

[0119]

[0120] The present invention is not limited by the embodiments described above and the attached drawings, but is intended to be limited by the appended claims. Accordingly, various substitutions, modifications, changes, and combinations of embodiments may be made by those skilled in the art without departing from the technical spirit of the invention as described in the claims, and such are also to be considered to fall within the scope of the present invention.

Claims

1. A data acquisition unit for acquiring SNP (Single Nucleotide Polymorphism) data from a blood sample; A data processing unit that obtains independent data for each SNP based on the above SNP data, obtains first interaction data between SNPs from the above SNP data through statistical analysis, and obtains second interaction data between SNPs from a gene network composed of genes corresponding to each of the above SNPs; A neural network processing unit that generates a graph-based neural network model by reflecting the independent data, the first interaction data, and the second interaction data, and trains the graph-based neural network model by adjusting the parameters of the graph-based neural network model; and Alzheimer's disease genetic risk diagnostic device comprising a diagnostic unit that diagnoses the genetic risk of Alzheimer's disease using a trained graph-based neural network model.

2. In Paragraph 1, The data processing unit performs an epistasis test to evaluate the statistical significance between SNPs based on the SNP data, thereby obtaining first interaction data between the SNPs, for an Alzheimer's disease genetic risk diagnosis device.

3. In Paragraph 1, The above data processing unit maps each of the above SNPs to a gene corresponding to each of the above SNPs, and obtains second interaction data between the above SNPs based on interaction data between the above genes in a gene network composed of the above genes, an Alzheimer's disease genetic risk diagnosis device.

4. In Paragraph 1, Alzheimer's disease genetic risk diagnosis device, wherein the neural network processing unit defines each SNP as a node of a graph-based neural network model and generates a graph-based neural network model by defining edges of the graph-based neural network model based on first interaction data between the SNPs and second interaction data between the SNPs.

5. In Paragraph 1, The above neural network processing unit trains the graph-based neural network model by adjusting a first parameter that determines the intensity of interaction data between SNPs, in an Alzheimer's disease genetic risk diagnosis device.

6. In Paragraph 5, The neural network processing unit trains the graph-based neural network model by adjusting a second parameter that determines the combination ratio of independent data for each of the SNPs, first interaction data between the SNPs, and second interaction data between the SNPs, in an Alzheimer's disease genetic risk diagnosis device.

7. In Paragraph 6, The neural network processing unit generates combined data in which independent data of each of the SNPs, first interaction data between the SNPs, and second interaction data between the SNPs are combined according to the second parameter, an Alzheimer's disease genetic risk diagnostic device.

8. In Paragraph 7, The neural network processing unit trains the graph-based neural network model by adjusting a third parameter for calculating the probability of genetic risk of Alzheimer's disease according to the combined data, an Alzheimer's disease genetic risk diagnostic device.

9. In Paragraph 8, The above diagnostic unit is an Alzheimer's disease genetic risk diagnostic device that calculates the genetic risk probability of the Alzheimer's disease using a learned graph-based neural network model and diagnoses the individual as a high-risk group for Alzheimer's disease if the calculated probability value exceeds a threshold value.

10. A step of obtaining SNP (Single Nucleotide Polymorphism) data from a blood sample; A step of obtaining independent data for each SNP based on the above SNP data; A step of obtaining first interaction data between SNPs from the above SNP data through statistical analysis; A step of obtaining second interaction data between SNPs from a gene network composed of genes corresponding to each of the above SNPs; A step of generating a graph-based neural network model by reflecting the independent data, the first interaction data, and the second interaction data; A step of training the graph-based neural network model by adjusting the parameters of the graph-based neural network model; and A method for diagnosing genetic risk of Alzheimer's disease, comprising the step of diagnosing genetic risk of Alzheimer's disease using a trained graph-based neural network model.

11. In Paragraph 10, A method for diagnosing genetic risk of Alzheimer's disease, wherein the step of obtaining the first interaction data comprises the step of obtaining the first interaction data between the SNPs by performing an epistasis test that evaluates the statistical significance between the SNPs based on the SNP data.

12. In Paragraph 10, A method for diagnosing genetic risk of Alzheimer's disease, comprising the step of obtaining the second interaction data, the step of mapping each of the SNPs to a gene corresponding to each of the SNPs; and the step of obtaining the second interaction data between the SNPs based on the interaction data between the genes in a gene network composed of the genes.

13. In Paragraph 10, A method for diagnosing genetic risk of Alzheimer's disease, comprising the step of generating the graph-based neural network model, the step of defining each SNP as a node of the graph-based neural network model; and the step of defining edges of the graph-based neural network model based on first interaction data between the SNPs and second interaction data between the SNPs.

14. In Paragraph 10, A method for diagnosing genetic risk of Alzheimer's disease, wherein the step of training the above-mentioned graph-based neural network model includes the step of adjusting a first parameter that determines the intensity of interaction data diffusion between SNPs.

15. In Paragraph 14, A method for diagnosing genetic risk of Alzheimer's disease, wherein the step of training the graph-based neural network model further includes the step of adjusting a second parameter that determines the combination ratio of independent data for each of the SNPs, first interaction data between the SNPs, and second interaction data between the SNPs.

16. In Paragraph 15, A method for diagnosing genetic risk of Alzheimer's disease, wherein the step of training the graph-based neural network model further comprises the step of generating combined data in which independent data for each of the SNPs, first interaction data between the SNPs, and second interaction data between the SNPs are combined according to the second parameter.

17. In Paragraph 16, A method for diagnosing genetic risk of Alzheimer's disease, wherein the step of training the graph-based neural network model further includes the step of adjusting a third parameter for calculating the probability of genetic risk of Alzheimer's disease according to the combined data.

18. In Paragraph 17, The step of diagnosing the genetic risk of Alzheimer's disease comprises: a step of calculating the probability of the genetic risk of Alzheimer's disease using a learned graph-based neural network model; and A method for diagnosing genetic risk of Alzheimer's disease, comprising a step of diagnosing a high-risk group for Alzheimer's disease when a calculated probability value exceeds a threshold value.

Images