Brain disease diagnosis model based on multiple peripheral tissue DNA methylation data

By combining deep neural networks and the LightGBM classifier in a brain disease diagnostic model, and utilizing DNA methylation data from multiple peripheral tissues, the limitations of single-source data in brain disease diagnosis have been overcome, achieving higher diagnostic accuracy and stability. This model is particularly suitable for the early diagnosis of epilepsy and Alzheimer's disease.

CN119480080BActive Publication Date: 2026-01-09DALIAN MARITIME UNIVERSITY
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411609793.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-11-12
Publication Date
2026-01-09
Estimated Expiration
2044-11-12

AI Technical Summary

Technical Problem

In existing technologies, peripheral tissue DNA methylation data from a single source ignores the complementarity and differences in methylation patterns between different biological samples in the diagnosis of brain diseases, resulting in reduced diagnostic accuracy and limited disease monitoring capabilities.

Method used

A diagnostic model for brain diseases based on DNA methylation data from various peripheral tissues was established. By combining deep neural networks (DNN) and LightGBM classifiers, high-level feature output vectors were obtained and classified using DNA methylation data from various peripheral tissues (such as peripheral blood, saliva, and oral cavity). The model was evaluated using mean accuracy, F1 score, and area under the ROC curve.

Benefits of technology

It improves the accuracy and stability of brain disease diagnosis. By leveraging the complementarity of DNA methylation data from various peripheral tissues, it enhances the reliability and precision of classification results, making it particularly suitable for the early diagnosis of complex diseases such as epilepsy and Alzheimer's disease.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119480080B_ABST
    Figure CN119480080B_ABST
Patent Text Reader

Abstract

The application discloses a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data, obtains a high-level feature output vector corresponding to the DNA methylation data of the peripheral tissue by using the preprocessed methylation data of the multiple peripheral tissues, and further obtains a classification result. The application uses multiple peripheral tissue DNA methylation data, which is easier to extract than single-source methylation data, and the multiple peripheral tissue DNA methylation data can be complementary. By using the DNA methylation data of multiple peripheral tissues of non-diseased persons and diseased persons, the stability and reliability of the classification result are improved. By combining the advantages of DNN in feature extraction and the high efficiency of the LightGBM classification algorithm in the classification task, the output result of the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data is more accurate.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical detection and diagnosis, and particularly relates to a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data. BACKGROUND

[0002] With the rapid development of genomics and epigenetics, DNA methylation as an important epigenetic regulation mechanism shows great potential in brain disease diagnosis. However, direct extraction of methylation data from brain tissue faces many challenges, including technical complexity and ethical issues of sampling, which limits its practical clinical application. The methylation data of peripheral tissues is not only easy to extract, but also has significant biological relevance between disease and brain tissue, thereby providing an important research approach for brain disease diagnosis. However, most disease diagnosis models only use methylation data from a single source (such as peripheral blood), ignoring the complementarity and difference of methylation patterns between different biological samples, which may lead to reduced diagnostic accuracy and limited disease monitoring and prediction capabilities. SUMMARY

[0003] The present application discloses a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data to overcome the above technical problems.

[0004] In order to achieve the above purpose, the technical scheme of the present application is:

[0005] A brain disease diagnosis model based on multiple peripheral tissue DNA methylation data, the establishment method of the model comprising the following steps:

[0006] S1: Obtain DNA methylation data of multiple peripheral tissues including non-diseased and diseased persons, and pre-process;

[0007] S2: Establish a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data, including multiple hidden layers and a prediction layer based on a LightGBM classifier;

[0008] S3: Based on the pre-processed DNA methylation data of the peripheral tissue, obtain a high-level feature output vector corresponding to the DNA methylation data of the peripheral tissue based on multiple hidden layers;

[0009] S4: According to the high-level feature output vector and the prediction layer based on the LightGBM classifier, obtain a classification result; and based on the average accuracy index, F1 score index and area under ROC curve index of the classification result, evaluate the classification result of the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data; to complete the establishment of the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data after passing the evaluation.

[0010] Furthermore, the DNA methylation data of the peripheral tissues includes DNA methylation data of peripheral blood, DNA methylation data of saliva, and DNA methylation data of the oral cavity.

[0011] Furthermore, the formula used to obtain the high-level feature output vector is as follows:

[0012]

[0013] In the formula: x represents the high-level feature output vector of the i-th sample; i A probe representing DNA methylation data of the peripheral tissue of the i-th sample; Dense M (x i ) represents the processing result of the dense layer on the DNA methylation data of the peripheral tissue of the i-th sample; Relu(·) represents the activation function; Dropout(·) represents the function used to prevent the model from overfitting; M represents the number of nodes in the hidden layer; i represents the patient sample index value.

[0014] Furthermore, the formula used to obtain the predicted diagnostic results for brain diseases is as follows:

[0015]

[0016] In the formula: represents the high-level feature output vector of the i-th sample; r is the index of the regression tree model in the LightGBM model; s is the total number of regression trees in the LightGBM model; f r Let F represent the output of the r-th independent regression tree model with leaf fractions; F represents the space of the classification regression tree.

[0017] The objective function of the output of the r-th independent regression tree model with leaf fractions is:

[0018]

[0019] In the formula: L(f r ) represents the objective function of the output of the r-th independent regression tree model with leaf fractions, i.e., the objective function that measures the predictive performance of the model; i represents the patient sample index value; n represents the total number of patient samples; Ω(f) represents a differentiable loss function; r represents the index of the regression tree in the LightGBM model; s represents the total number of regression trees in the LightGBM model; r ) is a regularization term;

[0020]

[0021] wherein: γ and λ are constants for controlling the degree of regularization; K is the number of leaf nodes in each regression tree model; w r is the value of the score on each leaf.

[0022] Further, the classification result of the brain disease includes: the number of true positives, the number of true negatives, the number of false positives, and the number of false negatives are obtained by using the DNA methylation data of multiple peripheral tissues of the non-diseased person and the DNA methylation data of multiple peripheral tissues of the diseased person.

[0023] Beneficial effects: the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data of the application inputs the preprocessed methylation data of multiple peripheral tissues into the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data combined with DNN and LightGBM classifier, obtains a high-level feature output vector corresponding to the DNA methylation data of the peripheral tissue, and then obtains a classification result. The application uses multiple peripheral tissue DNA methylation data, which is easier to extract than single-source methylation data, and the multiple peripheral tissue DNA methylation data can be complementary. By using the DNA methylation data of multiple peripheral tissues of non-diseased persons and diseased persons, the stability and reliability of the classification result are improved. By combining the advantages of DNN in feature extraction and the high efficiency of LightGBM classification algorithm in classification tasks, the output result of the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0025] Figure 1 A method flow chart for establishing the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data of the present application is provided.

[0026] Figure 2 A method flow chart for establishing the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data in the embodiments of the present application is provided.

[0027] Figure 3 A structure diagram of the disease diagnosis model based on DNN-LightGBM algorithm in the embodiments of the present application is provided. DETAILED DESCRIPTION

[0028] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0029] The present embodiment introduces a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data, as shown in Figure 1 and Figure 2 .

[0030] S1: Obtain DNA methylation data of multiple peripheral tissues and perform preprocessing; wherein the DNA methylation data of peripheral tissues includes DNA methylation data of peripheral blood, DNA methylation data of saliva and DNA methylation data of oral cavity;

[0031] Specifically, the present embodiment performs normalization and other preprocessing operations on the methylation data of multiple peripheral tissues of the non-diseased and patients, to provide high-quality data input for model training.

[0032] S2: Establish a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data, including multiple hidden layers and a prediction layer based on a LightGBM classifier; as shown in Figure 3 .

[0033] S3: Based on the preprocessed DNA methylation data of peripheral tissues, obtain a high-level feature output vector corresponding to the DNA methylation data of the peripheral tissues based on the multiple hidden layers, to learn the high-level feature representation in the DNA methylation data of each peripheral tissue;

[0034] Specifically, given a sample data set D = {(x i ,y i )}, wherein x i represents the probes of the DNA methylation data of peripheral tissues, y i is the corresponding binary label, 0 represents non-disease and 1 represents disease. Take x i as the input of the brain disease diagnosis model DNN based on multiple peripheral tissue DNA methylation data, and obtain the extracted high-level feature output vector

[0035] Preferably, the formula used to obtain the high-level feature output vector is as follows:

[0036]

[0037] wherein: represents the high-level feature output vector of the i-th sample; x i represents the probe of the DNA methylation data of the peripheral tissue of the i-th sample; Dense M (x i ) represents the processing result of the DNA methylation data of the peripheral tissue of the i-th sample by the dense layer; Relu(·) represents an activation function, which is a rectified linear unit that takes the output of the dense layer, maps all negative values to zero, and keeps positive values unchanged, thereby introducing a nonlinear feature; Dropout(·) represents a function for preventing model overfitting, which is an existing regularization technique for preventing model overfitting; M represents the number of nodes in the hidden layer; i represents the patient sample index value; in the embodiment, each sample includes multiple DNA methylation data of peripheral tissues, which can be optionally composed of DNA methylation data of peripheral blood, DNA methylation data of saliva, and DNA methylation data of oral cavity.

[0038] Specifically, M is the number of nodes in the hidden layer l (l = 1, 2…L), L is the total number of hidden layers, and l is the index number of the hidden layer. The dense layer extracts relevant patterns from the probe sequence of the DNA methylation data of the peripheral tissue of the sample and outputs the relevant patterns as features, wherein the relevant patterns are features and structures in the input data, such as the distribution and state of methylation sites. The rectified linear unit Relu(·) takes the output of the dense layer, maps all negative values to zero, and thereby introduces a nonlinear feature. Dropout is a technique for avoiding overfitting, in which some neurons and their connections are deleted (assigned to zero) during training.

[0039] S4: According to the high-level feature output vector and the prediction layer based on the LightGBM classifier, a LightGBM classification algorithm is used to obtain a classification result; and based on the average accuracy index, the F1 score index, and the area under the ROC curve index of the classification result, the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data is evaluated; and after the evaluation is qualified, the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data is established.

[0040] The classification result of the brain disease includes that the provider of the DNA methylation data of multiple peripheral tissues is a non-diseased person and the provider of the DNA methylation data of multiple peripheral tissues is a diseased person, and thereby the number of true positives, the number of true negatives, the number of false positives, and the number of false negatives can be obtained;

[0041] Specifically, the features extracted by the DNN and the labels are input into a LightGBM classifier. LightGBM is a highly efficient classification and regression tree ensemble learning algorithm, in which each step generates a basic regression tree model and adds it to the entire model, integrating their prediction results to improve classification accuracy. Assuming that the LightGBM model is composed of s regression trees, the optimization objective function is given by equation (2):

[0042]

[0043] wherein: represents the high-level feature output vector of the i-th sample; r is the number of regression tree models of the LightGBM model; s is the total number of regression trees of the LightGBM model; f r represents the output of the r-th independent regression tree model with leaf scores; F represents the space of the classification regression tree;

[0044] wherein, r = 1 represents the first tree, and f r corresponds to an independent regression tree with leaf scores. The objective function is defined by equation:

[0045]

[0046] wherein: L(f r ) represents the objective function for measuring the output of the r-th independent regression tree model with leaf scores, i.e., the objective function for measuring the prediction performance of the model; i represents the patient sample index value; n represents the total number of patient samples; represents the differentiable loss function; r represents the number of regression trees of the LightGBM model; s is the total number of regression trees of the LightGBM model; Ω(f r ) is the regularization term;

[0047] wherein, the differentiable loss function is used to measure the difference between the true label y i and the predicted label . The regularization term Ω(f r ), which can constrain the complexity of the model to avoid overfitting, is represented as:

[0048]

[0049] wherein: γ and λ are both constants for controlling the degree of regularization; K is the number of leaf nodes in each regression tree model; w r is the value of the score on each leaf;

[0050] Specifically, this embodiment uses 10-fold cross-validation to evaluate the performance of the classification algorithm. This process divides the dataset into 10 approximately equal folds. One fold serves as the test dataset, and the remaining nine folds are used as the training dataset. To optimize the hyperparameters of the classification model, a grid search method and 3-fold internal cross-validation are employed. This internal cross-validation is performed only on the training set. Finally, the model is fitted on the training set using the parameters that achieve the best performance, and then evaluated on the test set.

[0051] Specifically, this embodiment evaluates the performance of the classification model based on the average accuracy (ACC), F1 score (F1), and area under the ROC curve (AUC) of the classification results. In this embodiment, when all three indicators meet the requirements, it indicates that the model meets the usage requirements.

[0052] The formula for calculating the average accuracy index of the classification results is as follows:

[0053]

[0054] The formula for calculating the F1 score is as follows:

[0055]

[0056] In the formula: ACC represents the average accuracy of the classification results; TP represents the number of true positives; TN represents the number of true negatives; FP represents the number of false positives; FN represents the number of false negatives; and F1-score represents the F1 score.

[0057] The method for calculating the area under the ROC curve is as follows:

[0058] First, calculate the true positive rate and false positive rate using the following formulas:

[0059]

[0060] In the formula: TPR represents the true positive rate; FPR represents the false positive rate;

[0061] Then, the FPR and TPR values ​​corresponding to each threshold are used as coordinate points, and these points are plotted on a coordinate system. Connecting these points forms an ROC curve. The area under the ROC curve, i.e., the area under the curve between the true positive rate and the false positive rate, is then calculated.

[0062] Specifically, deep learning is a promising technology in machine learning, which attempts to create and simulate neural networks in the human brain to learn and interpret data. It has the potential to form abstract high-level representations by appropriately combining low-level features with the objective function, and high-level features are more abstract features further extracted from low-level features, which are more relevant to the semantic content of the disease, so as to find the rules of data. Therefore, feature selection is important. Among them, low-level features are closer to raw data, which are the methylation levels of individual CpG sites extracted directly from methylation data. These are the most basic methylation data, reflecting the methylation state of a specific position in the genome. The distinction between low-level features and high-level features lies in their level of abstraction and richness of semantic information. Low-level features are closer to raw data, while high-level features contain more abstract and semantic information, which can better represent and distinguish different disease states,

[0063] In this embodiment, we study the DNA methylation data of peripheral blood, saliva and oral cavity, and utilize the methylation information of multiple biological samples. Deep neural network (DNN) is used as a feature extractor to preprocess and extract features from multi-tissue methylation data. DNN can automatically learn high-level feature representations in data through multiple layers of nonlinear transformation, improving the generalization ability of the model. Then, the features extracted by DNN and the labels are input into the LightGBM classifier. Finally, a model for brain disease diagnosis using multi-tissue DNA methylation data is established.

[0064] The present embodiment fuses DNN with LightGBM algorithm, establishes a model, and proposes a binary classification brain disease diagnosis model using multi-tissue methylation data and combining LightGBM classifier with DNN architecture, i.e., a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data. First, the model automatically extracts high-level abstract features using DNN. Then, the high-level abstract features extracted by DNN are used as input of the LightGBM classifier together with class labels. In the DNN architecture, the LightGBM predictor is used instead of the neural network-based predictor to improve the accuracy of classification. The model combines the advantages of DNN in automatic feature extraction and complex pattern recognition, and the excellent performance of LightGBM in efficient classification and regression tasks, and uses multi-tissue methylation data from peripheral blood, saliva and oral cavity. By comprehensively analyzing multiple biological samples, the limitations of single data in complex disease diagnosis are overcome, and the stability and accuracy of diagnosis are significantly improved. The present embodiment can provide a more efficient and accurate new way for early detection and treatment of diseases. Taking epilepsy as an example, the model can use multi-tissue DNA methylation data of epilepsy patients and their healthy controls to perform efficient and accurate disease diagnosis. Through prediction analysis, doctors can identify potential epilepsy patients at an early stage, and provide strong support for subsequent disease progression monitoring and personalized treatment plan development. In addition, the model also opens up a new way for the diagnosis of other complex brain diseases (such as Alzheimer's disease), and predicts a broad application prospect in the field of precision medicine in the future.

[0065] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1.A brain disease diagnosis model based on a plurality of peripheral tissue DNA methylation data, characterized by, The model establishment method comprises the following steps: S1: Obtain DNA methylation data of multiple peripheral tissues including non-diseased and diseased persons, and pretreat; the DNA methylation data of the peripheral tissues includes DNA methylation data of peripheral blood, DNA methylation data of saliva and DNA methylation data of oral cavity; S2: Establish a brain disease diagnosis model based on multiple peripheral tissue DNA methylation data, comprising multiple hidden layers and a prediction layer based on a LightGBM classifier; S3: According to the pretreated DNA methylation data of the peripheral tissues, obtain a high-level feature output vector corresponding to the DNA methylation data of the peripheral tissues based on the multiple hidden layers; The formula used to obtain the high-level feature output vector is as follows: In the formula: represents the high-level feature output vector of the i-th sample; i represents the probe of the DNA methylation data of the peripheral tissue of the i-th sample; i represents the processing result of the dense layer on the DNA methylation data of the peripheral tissue of the i-th sample; i represents the activation function; represents the function for preventing model overfitting; M represents the number of nodes in the hidden layer; represents the patient sample index value;​​​ S4: According to the high-level feature output vector and the prediction layer based on the LightGBM classifier, obtain a classification result; and based on the average accuracy index, F1 score index and area under ROC curve index of the classification result, evaluate the classification result of the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data; after passing the evaluation, the brain disease diagnosis model based on multiple peripheral tissue DNA methylation data is established. 2.The brain disease diagnosis model based on multiple peripheral tissue DNA methylation data according to claim 1, wherein, The formula used to obtain the predicted diagnosis result of the brain disease is as follows: In the formula: represents the high-level feature output vector of the rth sample; i r is the number of regression tree models of the LightGBM model; s is the total number of regression trees of the LightGBM model; represents the output of the rth independent regression tree model with leaf score; F represents the space of the classification regression tree.​ Wherein, the target function of the output of the rth independent regression tree model with leaf score is: wherein: is a target function representing the output of the rth independent regression tree model with leaf fraction, i.e., a target function measuring the predictive performance of the model; is a patient sample index value; is a total number of patient samples; is a differentiable loss function; is a number of regression trees of the LightGBM model s is a total number of regression trees of the LightGBM model; is a regularization term; wherein: and are constants for controlling the degree of regularization; is the number of leaf nodes in each regression tree model; is the value of the score on each leaf. 3.The brain disease diagnosis model based on multiple peripheral tissue DNA methylation data according to claim 1, wherein, The classification result includes: the provider of the DNA methylation data of multiple peripheral tissues is non-diseased and the provider of the DNA methylation data of multiple peripheral tissues is diseased, according to the classification result, the number of true positives, the number of true negatives, the number of false positives and the number of false negatives can be obtained.

Citation Information

Patent Citations

  • Construction method and evaluation system of polycystic ovarian syndrome and anti-mullerian hormone prediction model based on peripheral blood DNA methylation

    CN116334212A

  • DNA methylation prediction method integrating multiple omics characteristics

    CN116758993A