Device for metabolic syndrome risk assessment of gut microbiota

Abstract

The application discloses a device for evaluating the risk of metabolic syndrome of intestinal flora, and the device comprises a preprocessing module, a screening module and a risk evaluation module. The preprocessing module is used for obtaining the relative abundance of each bacterial species in the intestinal flora after preprocessing the sequencing data corresponding to different batches of the intestinal flora. The screening module is used for introducing an empirical Bayes shrinkage estimation on the basis of a zero inflation model to obtain a correction model, screening after correcting the batch effect of the relative abundance by using the correction model, and obtaining the candidate features corresponding to each bacterial species. The risk evaluation module is used for evaluating the candidate features by using an evaluation model to obtain the risk result corresponding to the metabolic syndrome. The application can improve the accuracy of the risk evaluation of the metabolic syndrome.

Description

Technical Field

[0001] This application relates to, but is not limited to, the field of auxiliary diagnostic technology for metabolic syndrome, and in particular to a device for assessing the risk of metabolic syndrome based on gut microbiota. Background Technology

[0002] Metabolic syndrome is a complex clinical syndrome characterized by central obesity, hyperglycemia (prediabetes or diabetes), hypertension, and dyslipidemia (high triglycerides, low high-density lipoprotein cholesterol), and is a significant risk factor for cardiovascular and cerebrovascular diseases. Dysbiosis of the gut microbiota leads to impaired intestinal barrier function, allowing bacterial endotoxins such as lipopolysaccharides to enter the bloodstream, triggering low-grade systemic chronic inflammation, which in turn induces insulin resistance and vascular endothelial dysfunction. Furthermore, molecules produced by gut microbiota metabolism, such as short-chain fatty acids, branched-chain amino acids, and trimethylamine oxide, can directly or indirectly affect the host's insulin sensitivity, appetite regulation, hepatic lipid synthesis, and vascular tone, thereby synergistically regulating multiple metabolic indicators. Therefore, gut microbiota dysbiosis is one of the core links in the occurrence and development of metabolic syndrome. Treating the gut microbiota as a whole to assess an individual's risk of developing the four interrelated metabolic diseases—obesity, diabetes, hypertension, and hyperlipidemia—has a solid theoretical basis and significant clinical implications.

[0003] Existing technologies, such as bacterial 16S ribosomal gene amplicon sequencing, serve as a cost-effective and efficient technique for analyzing microbial composition. By amplifying and sequencing specific variable regions of bacterial genes, information on the species composition and relative abundance of the microbial community in a sample can be obtained. A model can then be trained on a training set corresponding to a single disease to obtain a single disease risk assessment result. However, amplicon sequencing data typically contains hundreds to thousands of microbial features, but most features are sparsely distributed in the sample, and batch effects exist between different sequencing batches. This results in the microbial features in the sequencing data failing to accurately reflect the actual situation, leading to inaccurate risk assessments for metabolic syndromes. Furthermore, risk assessments for metabolic syndromes often utilize classification models to assess the risk of a single metabolic disease, neglecting the fact that multiple metabolic syndromes often coexist and share risk factors, leading to omissions in the detection of existing metabolic diseases and consequently, inaccurate risk assessments for metabolic syndromes. Summary of the Invention

[0004] This application provides an apparatus for assessing the risk of metabolic syndrome based on gut microbiota, which can improve the accuracy of risk assessment for metabolic syndrome.

[0005] In a first aspect, embodiments of this application provide an apparatus for assessing the risk of metabolic syndrome based on gut microbiota, applied to an assessment system, comprising: The preprocessing module is used to preprocess the sequencing data corresponding to different batches of gut microbiota to obtain the relative abundance of each species in the gut microbiota. The screening module is used to introduce empirical Bayesian shrinkage estimation on the basis of the zero expansion model, correct for the batch effect of the relative abundance, and then screen to obtain candidate features corresponding to each species. The risk assessment module is used to evaluate the candidate features using an assessment model to obtain the risk results corresponding to each disease type of metabolic syndrome.

[0006] Secondly, an electronic device provided according to an embodiment of this application includes: At least one processor; At least one memory for storing at least one program; When at least one of the programs is executed by at least one of the processors, the apparatus implements the metabolic syndrome risk assessment of gut microbiota according to any one of the first aspects.

[0007] Thirdly, according to the embodiments of the application, a computer-readable storage medium is provided, storing computer-executable instructions for executing an apparatus for implementing a metabolic syndrome risk assessment of gut microbiota as described in any of the first aspects.

[0008] In summary, the apparatus for risk assessment of metabolic syndrome based on gut microbiota according to the above embodiments of this application includes: a preprocessing module, which is used to preprocess sequencing data corresponding to different batches of gut microbiota to obtain the relative abundance of each species in the gut microbiota; a screening module, which is used to introduce empirical Bayesian shrinkage estimation on the basis of a zero-expansion model, correct for the batch effect of relative abundance, and then screen to obtain candidate features corresponding to each species; and a risk assessment module, which is used to evaluate the candidate features using an assessment model to obtain the risk result corresponding to each disease type of metabolic syndrome. This application, based on the zero-inflation model closely matching actual data, introduces empirical Bayesian contraction estimation for modeling, thereby achieving stable estimation of the data. On this basis, it identifies and quantifies systematic biases between batches, accurately correcting batch effects for relative abundances with characteristics of zero inflation and excessive dispersion, and reasonably controlling the variation that batch effects may bring. After correction, it screens to obtain candidate features corresponding to each bacterial species. These candidate features can provide a good data foundation for subsequent metabolic syndrome. Then, reliable candidate features are input into the evaluation model for prediction, obtaining accurate risk results for each metabolic disease in metabolic syndrome. This approach can balance the accuracy of risk assessment for metabolic syndrome while avoiding omissions in the detection of metabolic diseases. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of an apparatus for assessing the risk of metabolic syndrome based on gut microbiota, according to one embodiment of this application. Figure 2 This is a flowchart of a metabolic syndrome risk assessment based on gut microbiota provided in one embodiment of this application; Figure 3 This is a schematic diagram of receiver operating characteristic (ROC) curves for training and independent validation sets of obesity, diabetes, hypertension, and hyperlipidemia, provided in one embodiment of this application. Figure 4 This application provides a probability density distribution map of risk scores for obesity, diabetes, hypertension, and hyperlipidemia in one embodiment. Figure 5 This is a hardware schematic diagram of an electronic device provided in one embodiment of this application. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0011] It is understandable that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, or the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0012] Metabolic syndrome is a complex clinical syndrome characterized by central obesity, hyperglycemia (prediabetes or diabetes), hypertension, and dyslipidemia (high triglycerides, low high-density lipoprotein cholesterol), and is a significant risk factor for cardiovascular and cerebrovascular diseases. Dysbiosis of the gut microbiota leads to impaired intestinal barrier function, allowing bacterial endotoxins such as lipopolysaccharides to enter the bloodstream, triggering low-grade systemic chronic inflammation, which in turn induces insulin resistance and vascular endothelial dysfunction. Furthermore, molecules produced by gut microbiota metabolism, such as short-chain fatty acids, branched-chain amino acids, and trimethylamine oxide, can directly or indirectly affect the host's insulin sensitivity, appetite regulation, hepatic lipid synthesis, and vascular tone, thereby synergistically regulating multiple metabolic indicators. Therefore, gut microbiota dysbiosis is one of the core links in the occurrence and development of metabolic syndrome. Treating the gut microbiota as a whole to assess an individual's risk of developing the four interrelated metabolic diseases—obesity, diabetes, hypertension, and hyperlipidemia—has a solid theoretical basis and significant clinical implications.

[0013] Existing technologies, such as bacterial 16S ribosomal gene amplicon sequencing, serve as a cost-effective and efficient technique for analyzing microbial composition. By amplifying and sequencing specific variable regions of bacterial genes, information on the species composition and relative abundance of the microbial community in a sample can be obtained. A model can then be trained on a training set corresponding to a single disease to obtain a single disease risk assessment result. However, amplicon sequencing data typically contains hundreds to thousands of microbial features, but most features are sparsely distributed in the sample, and batch effects exist between different sequencing batches. This results in the microbial features in the sequencing data failing to accurately reflect the actual situation, leading to inaccurate risk assessments for metabolic syndromes. Furthermore, risk assessments for metabolic syndromes often utilize classification models to assess the risk of a single metabolic disease, neglecting the fact that multiple metabolic syndromes often coexist and share risk factors, leading to omissions in the detection of existing metabolic diseases and consequently, inaccurate risk assessments for metabolic syndromes.

[0014] Based on this, embodiments of this application provide a device for assessing the risk of metabolic syndrome based on gut microbiota, which can improve the accuracy of risk assessment for metabolic syndrome. For example, this application introduces empirical Bayesian contraction estimation for modeling based on the zero-inflation model that fits the actual data, thereby achieving stable estimation of the data. On this basis, it identifies and quantifies systematic biases between batches, and can accurately correct batch effects for relative abundance with characteristics of zero inflation and excessive dispersion, reasonably control the variation that batch effects may bring, and screen after correction to obtain candidate features corresponding to each bacterial species. Candidate features can provide a good data foundation for subsequent metabolic syndrome. Then, reliable candidate features are input into the evaluation model for prediction to obtain accurate risk results corresponding to each metabolic disease of metabolic syndrome, thereby balancing the accuracy of risk assessment of metabolic syndrome while avoiding omissions in the detection of metabolic diseases.

[0015] A device for assessing the risk of metabolic syndrome from gut microbiota, applied to an assessment system, referenced Figure 1 and Figure 2 As shown, it includes: The preprocessing module is used to preprocess the sequencing data corresponding to different batches of gut microbiota to obtain the relative abundance of each species in the gut microbiota.

[0016] In some embodiments, the preprocessing module includes: a pre-processing module, which sequentially performs primer, quality control, noise reduction, and chimera removal preprocessing on the sequencing data to obtain feature sequences; and an abundance module, which calculates the relative abundance corresponding to the feature sequences.

[0017] For example, firstly, fecal samples are collected, and bacterial 16S ribosomal gene amplicon sequencing is performed on the fecal samples using a gene sequencer to obtain sequencing data. The sequencing data is saved in Fastq format, which is a text-based next-generation sequencing data file that stores biological sequences and their corresponding quality fractions. Next, these FASTQ files are imported into the microbiome analysis platform (Quantitative Insights Into Microbial Ecology 2, QIIME2) and converted into the platform's common data artifact format (i.e., .qza file) for subsequent unified processing and analysis. Then, since amplicon sequencing relies on PCR primers to target and amplify specific gene regions, the cutadapt tool integrated into the QIIME2 platform is used for primer removal. This ensures that all subsequent analyses are based on the target gene fragment itself, rather than artificially introduced primer sequences. The dada2 tool is then used to perform quality control, noise reduction, and chimera removal preprocessing operations, which can remove technical errors and interference from false positive species, generating characteristic sequences (amplicon sequence variants, ASV), providing a good data foundation for subsequent analyses.

[0018] For example, the generated feature sequences were annotated for species using the feature-classifier plugin on the QIIME2 platform. An amplicon database with a unified phylogenetic reference tree (such as Greengenes2) was selected as the reference species database, and the confidence threshold was set to 0.9 to ensure the reliability of the annotation results. A species abundance table was then generated. The rows of the species abundance table correspond to the sample names, the columns correspond to the bacterial species names at six taxonomic levels (phylum, class, order, family, genus to species), and the values ​​are the sequence read counts. These values ​​characterize the relative abundance of each species in the gut microbiota, with genus-level and species-level abundance data being the focus of this analysis.

[0019] Specifically, the relative abundance of each bacterial species ( ), Satisfy the following expression: ; in, This represents the number of bacterial species identified in the gut microbiota. This represents the total number of observed bacterial species in the gut microbiota. The relative abundance of each type of bacteria for each sample.

[0020] It is understandable that sequencing data can also be used to construct skin or oral cavity models using a general process, and then obtained through metagenomic detection, or it can be obtained from public datasets. This application does not limit the method for determining the sequencing data; it can be selected according to actual needs.

[0021] Specifically, the 16S rDNA sequencing public datasets of gut microbiota downloaded from databases such as NCBI (Center for Biotechnology Information) and NGDC (National Genomics Data Center) are shown in Table 1 below. Table 1 contains information on the item numbers of some of the downloaded datasets. The first column is the item number of the dataset in the database, the second column is the database source, the third column is the health or disease status, and the fourth column is the number of people in the dataset. To ensure data consistency, all data samples are of fecal type, the amplified regions are all V3-V4 regions (variable regions 3 and 4 of the 16S ribosomal gene sequence), and the sample collection locations are the same. Public datasets need to include sequencing Fastq and corresponding sample metadata tables. The sample metadata must include disease status labels for the samples, i.e., each sample is clearly labeled as healthy or suffering from one of the following: obesity, diabetes, hypertension, or hyperlipidemia.

[0022] Table 1

[0023] The screening module is used to introduce empirical Bayesian shrinkage estimation on the basis of the zero-inflation model, and after correcting for the batch effect of relative abundance, to screen and obtain the candidate features corresponding to each species.

[0024] Since sequencing data typically contains hundreds to thousands of microbial features, and most features are sparsely distributed in the sample, the number of zero values ​​(representing undetected species) in count data far exceeds the expected zero inflation of standard discrete distributions such as Poisson distribution. Simultaneously, the variance of the data is much greater than the excessive dispersion of the mean. In other words, sequencing data exhibits both zero inflation and excessive dispersion, resulting in the same characteristics being present in the relative abundance calculated based on the sequencing data. Existing technologies using Poisson or negative binomial regression models cannot properly handle zero inflation, leading to serious biases in parameter estimation. This application addresses the characteristics of zero inflation and excessive dispersion by using a zero inflation model to decompose the data generation process into two parts: whether zero values ​​occur and how many occur. This allows the modeling to better reflect the nature of the data.

[0025] During experiments, unavoidable variations in sequencing time, operators, reagent batches, or sequencing platforms introduce systematic variations unrelated to the biological variables of interest (such as disease states). This results in systematic noise superimposed on the inherent characteristics of the sequencing data, known as batch effects. Furthermore, in data from multiple independent batches, the sample size within each batch may be limited. Especially for certain rare bacterial species, directly estimating species effects (such as the strength of association with metabolic syndrome) based on small sample data from a single batch is highly unstable, with large variance in the estimated values, and easily influenced by batch-specific noise. This application's embodiments, based on a zero-inflation model that closely reflects reality, introduce empirical Bayesian contraction estimation for modeling. Using data from all batches and all bacterial species, a global, stable prior distribution is first estimated. Then, the initial estimates for each bacterial species-batch combination are contracted or adjusted towards this prior center, resulting in a more robust and reliable posterior estimate.

[0026] Thus, the embodiments of this application improve the accuracy and reliability of estimation by using posterior estimation and global information, identify and quantify systematic biases (position effect and scale effect) between batches, correct batches with small sample sizes or species with low detection frequency, and avoid under-correction or over-correction caused by high noise in single batch data. It can accurately correct batch effects for relative abundance with zero expansion and excessive dispersion characteristics, reasonably control the variation that batch effects may bring, and screen after correction to obtain candidate features corresponding to each species. The candidate features can provide a good data foundation for subsequent metabolic syndrome.

[0027] The risk assessment module is used to evaluate candidate features using an assessment model to obtain risk results for each disease type of metabolic syndrome.

[0028] For example, the evaluation model in this application embodiment is used to evaluate candidate features and predict the risk outcomes corresponding to metabolic diseases in metabolic syndrome, such as central obesity, hyperglycemia (prediabetes or diabetes), hypertension, and dyslipidemia (high triglycerides, low high-density lipoprotein cholesterol). The evaluation model can be a random forest model, gradient boosting tree, support vector machine, or deep neural network. Thus, this application embodiment inputs reliable candidate features into the evaluation model for prediction, obtaining accurate risk outcomes corresponding to each metabolic disease in metabolic syndrome, thereby balancing the accuracy of risk assessment for metabolic syndrome with avoiding omissions in the detection of metabolic diseases. This application embodiment does not limit the specific architecture of the evaluation model; a specific model can be selected according to actual needs.

[0029] In some embodiments, the screening module includes: a first calculation module, which is used to model the relative abundance of each batch to obtain a normal distribution; a second calculation module, which is used to model the prior mean effect and prior variability effect corresponding to the relative abundance of each batch through probability statistics, and to model the prior mean effect and prior variability effect in combination with the normal distribution using Bayes' theorem to obtain the observed readings corresponding to the relative abundance; and a third calculation module, which is used to screen the observed readings after correction to obtain candidate features.

[0030] For example, considering that the abundance data contains a large number of zero values, the non-zero relative abundance is modeled on a logarithmic scale. After taking the logarithm, the non-zero relative abundance follows a normal distribution, where the mean of the normal distribution is affected by biological covariates and batch effects, and the variance is affected by batch heterogeneity. Based on this, combined with empirical Bayesian contraction estimation, an initial prior mean effect and prior variability effect are calculated for each combination of relative abundance and batch using frequency statistics methods. This establishes the prior distribution. Then, Bayes' theorem is used to combine the prior mean effect and prior variability effect in a small sample of this batch with the prior distribution (i.e., the normal distribution) to model the observation readings, obtaining posterior estimates (i.e., estimates of the mean effect, estimates of the variability effect, and estimates of the standard deviation).

[0031] Specifically, the observed readings satisfy the following expression: ; in, This represents the observed reading of the relative abundance of the j-th fecal sample in the i-th batch. As biologically significant covariates, The coefficient representing the influence of biological covariates on relative abundance. Let be the prior average effect of the i-th batch on relative abundance. Let be the prior variability effect of the i-th batch on the relative abundance. The total standard deviation of relative abundance. The error is random and follows a standard normal distribution. It is a binary indicator variable used to evaluate zero-value acceptance and rejection, indicating whether the observed zero value is a biologically non-existent zero or a technical zero that was not detected due to technical reasons such as insufficient sequencing depth.

[0032] In some embodiments, the third calculation module includes: a correction module, which is used to correct the observed readings by means of the biological covariates corresponding to the normal distribution, the estimated average effect, the estimated variability effect, and the estimated standard deviation, to obtain abundance features; and a selection module, which is used to screen the abundance features by means of linear discriminant analysis effect size to obtain candidate features.

[0033] For example, the logarithm of the observations is taken, the estimated covariate effects and the contracted batch location effects are subtracted, and the contracted batch-scale effect is divided to standardize the variance between different batches. The desired covariate effects are then added back, the exponential transformation is performed back to the original scale, and the result is multiplied by a zero indicator to keep the value at zero. Finally, the corrected values ​​of all observations for each sample are re-standardized so that the total sequencing depth for each sample remains consistent with that before correction.

[0034] Specifically, the abundance feature satisfies the following expression: ; in, Let be the abundance feature of the relative abundance of the j-th sample in the i-th batch. For biological covariate effects, This is due to the batch position effect after shrinkage. This is the batch scaling effect after shrinkage (i.e., the estimated variability effect). It is a binary indicator variable.

[0035] In some embodiments, the selection module includes: a preliminary screening module, which performs a Kruskal-Wallis test between the disease group and the healthy control group, and then uses Wilcoxon to screen between the disease group and the healthy control group to obtain preliminary screening results, wherein the disease group consists of data associated with disease labels in the abundance features, and the healthy control group consists of data associated with health labels in the abundance features; and a decision module, which scores the preliminary screening results using linear discriminant analysis to obtain a linear discriminant score, and determines candidate features based on the linear discriminant score and the abundance features.

[0036] In some embodiments, the decision module includes: a frequency statistics module, which is used to calculate the frequency value of the abundance feature; and a target decision module, which is used to screen candidate features from the abundance features whose frequency value is less than a first threshold and whose linear discriminant score is greater than a second threshold.

[0037] For example, for each bacterial species or taxonomic level, the linear discriminant analysis effect size (LEFSE) is used to calculate the statistical significance of differences in abundance characteristics between the disease group and the healthy group. Taking the disease group and the healthy control group as examples, the Kruskal-Wallis test is first used to screen for differentially expressed microorganisms between the two groups, then the Wilcoxon test is used to verify intragroup consistency, and finally, linear discriminant scoring is used to assess the influence of differentially expressed microorganisms between the two groups. Candidate features with a frequency value less than a first threshold and a linear discriminant score greater than a second threshold are screened to find microbial biomarkers that have both statistical significance and influence.

[0038] In some embodiments, the risk assessment module includes: a construction module, which is used to construct assessment sub-models corresponding to the disease types of metabolic syndrome respectively, and to construct an assessment model based on all the assessment sub-models; a prediction module, which is used to input candidate features into each assessment sub-model in the assessment model respectively to obtain the risk value corresponding to the disease type; and a mapping module, which is used to map the risk value to a preset risk level to obtain the risk result.

[0039] Understandably, gradient boosting trees are prone to overfitting due to their complex parameters, support vector machines are inefficient with a large number of features, and neural networks require a large amount of data to realize their advantages and have poor interpretability. Considering the practical conditions of multi-source data, moderate sample size, multi-task output, and the need for strong interpretability in gut microbiota, this application selects random forest as the evaluation model to achieve a balance between accuracy, stability, efficiency, and interpretability.

[0040] For example, refer to Figure 3 As shown, this application constructs a separate evaluation sub-model with a random forest architecture for each disease type in metabolic syndrome. Then, the evaluation sub-models are constructed in parallel to form an evaluation model. Candidate features are input into the evaluation sub-models respectively. Each evaluation sub-model outputs the risk value (range 0-100) of the sample belonging to the disease category. The risk value is then mapped to a preset risk level (low risk, medium risk, high risk). The risk level is divided according to the probability density distribution of risk scores of disease groups and healthy groups in public databases. The area where healthy subjects are concentrated is classified as the low-risk range, the area where disease patients are concentrated is classified as the high-risk range, and the area where the two overlap is classified as the medium-risk range. Finally, a comprehensive multi-disease risk report is generated as the risk result. The final statistics are shown in Table 2 below.

[0041] Table 2

[0042] For example, after obtaining the continuous risk values ​​output, they need to be transformed into more clinically relevant classification information. To this end, thresholds for low, medium, and high risk levels are established based on the probability density distributions of risk values ​​from a large number of known disease and healthy samples in a public database. Specifically, the core region of risk distribution in the healthy population is defined as the low-risk range, the core region of risk distribution in the diseased population is defined as the high-risk range, and the overlapping region where the two distributions intersect is defined as the medium-risk range. First, the model is tested using an independent public database different from the training data source to verify its generalization ability to the new data cohort. Then, it is validated through a prospective, small-scale real-world study: fecal samples from 20 individuals are collected, standard 16S rDNA amplicon sequencing is performed, species abundance data is constructed, and then input into the model for risk scoring and level assessment.

[0043] Specifically, such as Figure 4 As shown, the risk scores of healthy subjects are concentrated between 0 and 50, while the risk scores of patients with diseases are mostly between 75 and 100. Based on the probability density function distribution diagrams of the four diseases mentioned above, the risk levels are divided as follows: low risk, ≤50; medium risk, (50, 75); high risk, (75, 100). If a person's gut microbiota abundance data outputs the following risk scores after model evaluation: obesity 55, diabetes 85, hypertension 25, hyperlipidemia 95, the following report will be generated: This subject has medium risk for obesity, high risk for diabetes, low risk for hypertension, and high risk for hyperlipidemia.

[0044] In some embodiments, the construction module includes: a data acquisition module, which is used to acquire training sets and validation sets corresponding to all disease types of metabolic syndrome; a parameter determination module, which is used to determine the number of decision trees and the number of node independent variables in the training set through cross-validation; and an optimization module, which is used to construct a random forest model with the number of decision trees and the number of node independent variables, and to validate and optimize the random forest model on the validation set based on the maximum Yoden index to obtain an evaluation sub-model.

[0045] When constructing a random forest model for each disease, the dataset is randomly divided into 5 equal parts. In each iteration, one part is used as the validation set, and the remaining 4 parts are used as the training set. A strategy combining grid search and K-fold cross-validation is used to determine the number of decision trees (i.e., the total number of decision trees) and the number of node independent variables (i.e., the maximum depth of each tree) in the random forest. In each round of cross-validation evaluation, the negative mean squared error (NSF) is used as the performance metric. Maximizing the performance metric effectively minimizes the NSF, thus selecting the optimal hyperparameter configuration, i.e., the optimal number of decision trees and node independent variables. After determining the optimal number of decision trees and node independent variables, the model is fitted on the entire training set using the optimal number of decision trees and node independent variables to obtain the random forest model. To find the optimal threshold for distinguishing between positive (true positive rate, i.e., sensitivity, TPR) and negative (true negative rate, i.e., specificity, FDR) on the receiver operating characteristic curve, this application selects the maximum Youden's Index (J) as the optimal balance point to validate and optimize the random forest model on the validation set. Table 2 is constructed based on the sensitivity and specificity of different evaluation sub-models on the training set, as shown below. Table 2 gives the sensitivity and specificity of the optimal cutoff value for different evaluation sub-models. TPR, FDR, and J satisfy the following expressions: ; Wherein, TP represents the number of true positives, FN represents the number of false negatives, FP represents the number of false positives, J represents the maximum Yoden index, TPR represents the true positive rate, and FDR represents the true negative rate.

[0046] Table 3

[0047] It is understood that the above embodiments of this application have the following beneficial effects: (I) Significantly Improved Assessment Accuracy and Robustness: Through rigorous data standardization and screening processes, discriminative features are extracted from limited species information, thereby constructing a risk assessment model with stronger generalization ability. Specifically, the predictive AUC (Area Under Receiver Operating Characteristic) for the hypertension independent validation set is 0.989, with a sensitivity of 98.7% and a specificity of 95.6%, superior to patent CN202411292904.6; the predictive AUC for the diabetes independent validation set is 0.961, with a sensitivity of 97.6% and a specificity of 97.8%, superior to patent CN202310398006.8. (II) Enhanced clinical applicability and interpretability: The quantitative risk scores and clear risk levels provided offer richer and more intuitive decision-making information compared to simple binary classification results (CN107480474B), facilitating individualized health management, risk stratification, and early intervention. (III) High assessment efficiency: Unlike other patents (CN115472231B, CN119736383B, CN107480474B) which only assess a single disease, this application directly obtains a comprehensive risk assessment of four major metabolic diseases, avoiding the tediousness of multiple analyses and improving assessment efficiency and ease of use.

[0048] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned apparatus for assessing the metabolic syndrome risk of gut microbiota. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0049] Please see Figure 5 , Figure 5 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 501 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 502 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 502 and is called by the processor 501 to execute the apparatus for assessing the risk of metabolic syndrome based on gut microbiota in the embodiments of this application. The input / output interface 503 is used to implement information input and output; The communication interface 504 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 505 transmits information between various components of the device (e.g., processor 501, memory 502, input / output interface 503, and communication interface 504); The processor 501, memory 502, input / output interface 503, and communication interface 504 are connected to each other within the device via bus 505.

[0050] In some embodiments, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned apparatus for assessing the metabolic syndrome risk of gut microbiota.

[0051] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0052] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0053] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0054] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0055] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0056] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0057] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0058] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0059] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0060] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0061] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A device for assessing the risk of metabolic syndrome based on gut microbiota, characterized in that, Applied to evaluation systems, including: The preprocessing module is used to preprocess the sequencing data corresponding to different batches of gut microbiota to obtain the relative abundance of each species in the gut microbiota. The screening module is used to introduce empirical Bayesian shrinkage estimation on the basis of the zero expansion model, correct for the batch effect of the relative abundance, and then screen to obtain candidate features corresponding to each species. The risk assessment module is used to evaluate the candidate features using an assessment model to obtain the risk results corresponding to each disease type of metabolic syndrome.

2. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 1, characterized in that, The filtering module includes: The first calculation module is used to model the relative abundance of each batch to obtain a normal distribution; The second calculation module is used to calculate the prior average effect and prior variability effect corresponding to the relative abundance of each batch through probability statistics, and to model the prior average effect and prior variability effect in combination with the normal distribution using Bayes' theorem to obtain the observation readings corresponding to the relative abundance. The third calculation module is used to correct and filter the observed readings to obtain the candidate features.

3. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 2, characterized in that, The third calculation module includes: The correction module is used to correct the observation readings using the biological covariates corresponding to the normal distribution, the estimated average effect, the estimated variability effect, and the estimated standard deviation, to obtain the abundance characteristics. The selection module is used to filter the abundance features by linear discriminant analysis effect size to obtain candidate features.

4. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 3, characterized in that, The selection module includes: The initial screening module is used to perform Kruskal-Wallis test between the disease group and the healthy control group, and then combine Wilcoxon to screen the disease group and the healthy control group respectively to obtain the initial screening results. The disease group is the data associated with the disease label in the abundance feature, and the healthy control group is the data associated with the health label in the abundance feature. The decision module is used to score the initial screening results by linear discriminant analysis to obtain a linear discriminant score, and select the candidate features from the abundance features based on the linear discriminant score.

5. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 4, characterized in that, The decision-making module includes: A frequency statistics module, which is used to calculate the frequency values ​​of the abundance features; The target decision module is used to filter candidate features from the abundance features whose frequency value is less than a first threshold and whose linear discriminant score is greater than a second threshold.

6. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 1, characterized in that, The risk assessment module includes: A construction module is used to construct assessment sub-models corresponding to the disease types of metabolic syndrome, and to construct the assessment model based on all the assessment sub-models; The prediction module is used to input the candidate features into each of the evaluation sub-models in the evaluation model to obtain the risk value corresponding to the disease type; A mapping module is used to map the risk value to a preset risk level to obtain the risk result.

7. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 6, characterized in that, The construction module includes: The data acquisition module is used to acquire training sets and validation sets corresponding to all disease types of metabolic syndrome. A parameter determination module is used to determine the number of decision trees and the number of node independent variables in the training set through cross-validation. An optimization module is used to construct a random forest model using the number of decision trees and the number of node independent variables, and to validate and optimize the random forest model on the validation set based on the maximum Yangon index, thereby obtaining the evaluation sub-model.

8. The apparatus for assessing the risk of metabolic syndrome based on gut microbiota according to claim 1, characterized in that, The preprocessing module includes: The preprocessing module is used to sequentially preprocess the sequencing data by primer selection, quality control, noise reduction, and chimera removal to obtain the characteristic sequence. An abundance module is used to calculate the relative abundance corresponding to the feature sequence.

9. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; when at least one of the programs is executed by at least one of the processors, the means for implementing the metabolic syndrome risk assessment of gut microbiota as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are used to perform a metabolic syndrome risk assessment of the gut microbiota according to any one of claims 1 to 8.

Images