Obesity risk prediction report generation system and method based on knowledge base

Abstract

The invention relates to the technical field of medical data privacy protection and information security, in particular to an obesity risk prediction report generation system and method based on a knowledge base, and the system comprises a data collection module, a risk prediction model module, a knowledge base module and a report generation module. According to the method, multi-dimensional data of intestinal microorganisms, gene detection, dietary habits, lifestyles and the like of the user are analyzed, a knowledge base and machine learning model combination technology is utilized, a personalized obesity risk assessment report can be generated, and therefore multi-dimensional and accurate obesity risk assessment is achieved.

Description

Technical Field

[0001] The present invention relates to the field of precision medicine technology, and in particular to a system and method for generating an obesity risk prediction report based on a knowledge base, and more particularly to a system and method for generating a personalized obesity risk prediction report based on a machine learning algorithm by combining multidimensional data of intestinal microbial detection, genes, diet and lifestyle. Background Art

[0002] Obesity is a complex metabolic disease that involves the interaction of multiple factors such as genetics, environment, microorganisms, and lifestyle. Gut microorganisms play a key role in the metabolic regulation of the host. By analyzing the gut microbial community and its relationship with genes, diet, and lifestyle, we can effectively assess an individual's obesity risk and propose personalized intervention plans. However, current existing technologies usually only focus on data in one dimension (such as genes or lifestyle) and cannot achieve multi-dimensional and accurate obesity risk assessment.

[0003] To this end, a knowledge base-based obesity risk prediction report generation system and method are proposed. Summary of the invention

[0004] Based on this, it is necessary to provide a knowledge base-based obesity risk prediction report generation system and method to address the above technical issues.

[0005] According to a first aspect of the present invention, a knowledge base-based obesity risk prediction report generation system is provided, comprising: a data acquisition module, for acquiring and preprocessing multidimensional evaluation data of a target user to obtain a multidimensional evaluation preprocessing result, wherein the multidimensional evaluation data includes initial intestinal microbial sequencing data, initial gene detection data, dietary habit information, and lifestyle information; the multidimensional evaluation preprocessing result includes intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores, and lifestyle scores; a risk prediction model module, for inputting the multidimensional evaluation preprocessing result into a trained multi-classification obesity risk prediction model, outputting the probability distribution of the target user belonging to high risk, medium risk, and low risk, and calculating the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtaining the risk classification level result to which the target user belongs according to the mapping relationship formed between the comprehensive obesity risk score and the risk classification level field; a knowledge base module , used to construct a microbial database, a gene database, a diet database and a lifestyle database, and based on the multi-dimensional evaluation data of the target user, screen out multi-dimensional risk assessment features from the microbial database, the gene database, the diet database and the lifestyle database, and use a pattern matching algorithm to perform correlation analysis on the multi-dimensional risk assessment features and the key features related to obesity risk to obtain the correlation analysis results, and based on the correlation analysis results, screen out the key risk impact features related to obesity according to the evidence-based weights, and store the corresponding health management advice information based on the risk classification level results; a report generation module is used to generate a personalized obesity risk assessment report based on the comprehensive obesity risk score of the target user, the risk classification level results to which the target user belongs, the key risk impact features related to obesity, the evidence-based medical information corresponding to the key risk impact features related to obesity, and the health management advice information corresponding to the risk classification level results to which the target user belongs.

[0006] Optionally, the data collection module includes: an intestinal microorganism collection and processing submodule, which is used to collect fecal samples of target users, extract microbial DNA from the fecal samples, obtain initial intestinal microorganism sequencing data through 16S rRNA sequencing or metagenomic sequencing, and perform quality control and sequence alignment processing, species classification analysis and abundance analysis, functional annotation and pathway analysis, data standardization and dimensionality reduction processing and feature screening processing on the initial intestinal microorganism sequencing data to screen out intestinal microorganism assessment-related features; a gene collection and processing submodule, which is used to upload initial gene detection data in multiple data formats provided by a third-party gene detection platform, and perform format conversion, field mapping processing and genetic feature extraction on the initial gene detection data to screen out SNP sites that are significantly associated with obesity, wherein the SNP site that is significantly associated with obesity refers to a SNP site whose β value or OR value associated with obesity is greater than the first type of preset value, and the polygenic risk score value is calculated through the individual polygenic risk score calculation formula; a lifestyle and diet data collection and processing submodule, which is used to input dietary habit information and lifestyle information, and obtain dietary habit scores and lifestyle scores through weighted integration processing.

[0007] Optionally, the intestinal microorganism collection and processing submodule includes: an intestinal microorganism collection submodule, which is used to collect fecal samples of target users, extract microbial DNA from the fecal samples, and obtain initial intestinal microorganism sequencing data through 16S rRNA sequencing or metagenomic sequencing; a quality control and sequence alignment submodule, which is used to check the quality of intestinal microorganism sequencing data, filter out low-quality sequences and connector sequences contained in the initial intestinal microorganism sequencing data, and obtain intestinal microorganism sequencing processing data, wherein the low-quality sequence refers to a sequence quality score less than the second category preset value; and perform sequence alignment processing on the intestinal microorganism sequencing processing data and the reference microbial genome database to obtain intestinal microorganism sequencing alignment result data; a species classification and abundance analysis submodule, which is used to perform species classification on the intestinal microorganism sequencing alignment result data to obtain species information and abundance information of the intestinal microbial community; a functional annotation and pathway analysis submodule , which is used to perform functional prediction on the species and abundance information of intestinal microorganisms according to the gene function database, obtain the functional potential information of intestinal microorganisms, and extract the functional pathway characteristics related to obesity based on the functional potential information of intestinal microorganisms; the data standardization and dimensionality reduction submodule is used to standardize the abundance information and functional pathway characteristics of intestinal microorganisms to obtain standardized intestinal microorganism data, and perform dimensionality reduction on the standardized intestinal microorganism data to extract the effective features of intestinal microorganisms; the feature screening submodule is used to call the microbial database of the knowledge base module, and use the abundance, species and functional pathway characteristics of intestinal microorganisms as constraints to screen out intestinal microorganism evaluation-related features from the effective features of intestinal microorganisms.

[0008] Optionally, the gene collection and processing submodule includes: a gene data collection submodule, which is used to upload gene initial detection data in multiple data formats provided by a third-party gene detection platform; a format conversion submodule, which is used to convert the format of the gene initial detection data in multiple data formats based on a rule-based parsing algorithm to obtain gene detection data in JSON format; a field mapping processing submodule, which is used to construct a field mapping table, and based on the constructed field mapping table, perform field mapping on the gene detection data to obtain gene detection mapping data; a genetic feature extraction submodule, which is used to screen the SNP sites contained in the gene detection mapping data by calling the gene database of the knowledge base module, and screen out the SNP sites that are significantly associated with obesity, wherein the SNP sites that are significantly associated with obesity refer to the SNP sites whose β value or OR value associated with obesity is greater than the first type preset value, and the polygenic risk score value is calculated by the individual polygenic risk score calculation formula.

[0009] Optionally, the individual's polygenic risk score calculation formula is: ; In the formula, PRS represents the polygenic risk score, Gj represents the number of allele copies of an individual at the j-th SNP site, Gj takes one of 0, 1, and 2, Rj represents the β value or OR value associated with the j-th SNP site and obesity, and m represents the number of SNP sites used for calculation after screening.

[0010] Optionally, the lifestyle and diet data collection and processing submodule includes: a lifestyle and diet data input submodule, used to input dietary habit information and lifestyle information, wherein the dietary habit information includes the frequency and amount of intake of sugary foods, the frequency and amount of intake of high-fat foods, and fat source information, and the lifestyle information includes exercise information and sleep information; an integration processing submodule, used to use weighted integration technology to perform weighted integration processing on the input dietary habit information and lifestyle information to generate dietary habit scores and lifestyle scores.

[0011] Optionally, the risk prediction model module includes: a data processing and integration submodule, which is used to obtain historical multi-dimensional evaluation preprocessing results and historical actual obesity risk values ​​of multiple users, standardize and feature encode the historical multi-dimensional evaluation preprocessing results, and obtain historical multi-dimensional standardized feature vectors, wherein the historical multi-dimensional standardized feature vectors include intestinal microbial data standardized feature vectors, gene data standardized feature vectors, dietary habit data standardized feature vectors, and lifestyle data standardized feature vectors; a feature selection submodule, which is used to combine the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, and call the microbial database, gene database, dietary database, and lifestyle database of the knowledge base module to screen out key features related to obesity risk; a model construction and training submodule, which is used to select key features related to obesity risk based on the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, and call the microbial database, gene database, dietary database, and lifestyle database of the knowledge base module. According to the screened key features related to obesity risk, a multi-classification obesity risk prediction model is constructed and trained based on the Softmax regression algorithm. The multi-classification obesity risk prediction model is used to calculate the probability distribution of each user belonging to high risk, medium risk, and low risk; the risk score calculation and classification submodule is used to input the multi-dimensional evaluation preprocessing results of the target user into the trained multi-classification obesity risk prediction model, output the probability distribution of the target user belonging to high risk, medium risk, and low risk, and calculate the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtain the risk classification level result of the target user according to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field; the model update and optimization submodule is used to adopt incremental learning technology to continuously update and optimize the multi-classification obesity risk prediction model.

[0012] Optionally, the feature selection submodule includes: a feature matrix construction submodule, which is used to construct a feature matrix based on the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, with each user as a row and the historical multi-dimensional standardized feature vector of each user as a column; a target variable preparation submodule, which is used to take the historical actual obesity risk value of each user as the target variable; and a Lasso regression modeling submodule, which is used to establish a regularized inverse Lasso regression objective function based on the feature matrix and the target variable, wherein the Lasso regression objective function is: ; In the formula, yi represents the user's actual historical obesity risk value, i represents the predicted obesity risk value, βi represents the regression coefficient of the historical multidimensional standardized feature vector i, λ represents the regularization parameter, which is used to control the strength of L1 regularization, and the regularization parameter λ is selected by cross-validation, and n represents the total number of users; a feature screening submodule is used to screen out features that are significantly correlated with obesity risk based on L1 regularization of the Lasso regression objective function, wherein the features that are significantly correlated with obesity risk refer to the features that are determined by the value of βi approaching zero when the L1 regularization of the Lasso regression objective function is performed, and the microbial database, gene database, diet database and lifestyle database of the knowledge base module are called to verify the biological significance of the features that are significantly correlated with obesity risk, and the features that have passed the biological significance verification are used as key features related to obesity risk.

[0013] Optionally, the knowledge base module includes: a microbial database for storing intestinal microbial information related to obesity and evidence information corresponding to the intestinal microbial information; a gene database for recording gene variations directly related to obesity and evidence information of the interaction between gene variations and microorganisms; a diet database for recording information and evidence information on the impact of different dietary patterns on obesity, and information and evidence information on the regulatory effects of different dietary patterns on intestinal microorganisms; a lifestyle database for recording information and evidence information on the impact of different lifestyles on obesity, and information and evidence information on the regulatory effects of different lifestyles on intestinal microorganisms; an association analysis submodule for screening out multidimensional risk assessment features from the microbial database, gene database, diet database and lifestyle database based on the multidimensional evaluation data of the target user, and using a pattern matching algorithm to perform correlation analysis on the multidimensional risk assessment features and key features related to obesity risk to obtain correlation analysis results, and based on the correlation analysis results and according to the evidence-based weights, screen out key risk impact features related to obesity; a health management advice storage submodule for storing corresponding health management advice information based on the risk classification level results.

[0014] According to a second aspect of the present invention, a method for generating an obesity risk prediction report based on a knowledge base is provided, comprising: collecting and preprocessing multidimensional evaluation data of a target user through a data acquisition module to obtain a multidimensional evaluation preprocessing result, wherein the multidimensional evaluation data includes initial intestinal microbial sequencing data, initial gene detection data, dietary habit information, and lifestyle information; the multidimensional evaluation preprocessing result includes intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores, and lifestyle scores; inputting the multidimensional evaluation preprocessing result into a trained multi-classification obesity risk prediction model through a risk prediction model module, outputting a probability distribution of whether the target user belongs to high risk, medium risk, and low risk, and calculating a comprehensive obesity risk score of the target user based on a comprehensive risk score calculation formula, and obtaining a risk classification level result to which the target user belongs according to a mapping relationship formed between the comprehensive obesity risk score and the risk classification level field; and obtaining a risk classification level result to which the target user belongs through a knowledge base. The module constructs a microbial database, a gene database, a dietary database and a lifestyle database. Based on the multi-dimensional evaluation data of the target user, multi-dimensional risk assessment features are screened out from the microbial database, the gene database, the dietary database and the lifestyle database. The pattern matching algorithm is used to perform correlation analysis on the multi-dimensional risk assessment features and the key features related to obesity risk to obtain the correlation analysis results. Based on the correlation analysis results, according to the evidence-based weights, the key risk impact features related to obesity are screened out, and based on the risk classification level results, the corresponding health management advice information is stored. The report generation module generates a personalized obesity risk assessment report based on the comprehensive obesity risk score of the target user, the risk classification level results of the target user, the key risk impact features related to obesity, the evidence-based medical information corresponding to the key risk impact features related to obesity, and the health management advice information corresponding to the risk classification level results of the target user.

[0015] The present application provides a knowledge base-based obesity risk prediction report generation system and method. The system analyzes multi-dimensional data such as the user's intestinal microorganisms, genetic testing, eating habits and lifestyle, and uses a combination of knowledge base and machine learning model technology to generate a personalized obesity risk assessment report, thereby achieving multi-dimensional and accurate obesity risk assessment and providing health intervention recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a system principle block diagram of the present invention.

[0017] Figure 2 This is a system principle block diagram of the data acquisition module of the present invention.

[0018] Figure 3 This is a system principle block diagram of the intestinal microorganism collection and processing submodule of the present invention.

[0019] Figure 4 This is a system principle block diagram of the gene collection and processing submodule of the present invention.

[0020] Figure 5 This is a system principle block diagram of the lifestyle and diet data collection and processing submodule of the present invention.

[0021] Figure 6 This is a system principle block diagram of the risk prediction model module of the present invention.

[0022] Figure 7 This is a system principle block diagram of the feature selection submodule of the present invention.

[0023] Figure 8 This is a system principle block diagram of the knowledge base module of the present invention.

[0024] Fig. 9 It is the overall flow chart of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below through specific implementation methods in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0026] Embodiment 1

[0027] Reference Figure 1-8 , an obesity risk prediction report generation system based on a knowledge base, the system includes a data acquisition module, a risk prediction model module, a knowledge base module and a report generation module.

[0028] In some embodiments, the data collection module is used to collect and preprocess multi-dimensional evaluation data of the target user to obtain a multi-dimensional evaluation preprocessing result, wherein the multi-dimensional evaluation data includes initial intestinal microbial sequencing data, initial gene detection data, dietary habit information and lifestyle information; the multi-dimensional evaluation preprocessing result includes intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores and lifestyle scores.

[0029] For further information, see Attachment Figure 2The data collection module includes: an intestinal microorganism collection and processing submodule, which is used to collect fecal samples of target users, extract microbial DNA from fecal samples, obtain initial intestinal microorganism sequencing data through 16S rRNA sequencing or metagenomic sequencing, and perform quality control and sequence alignment processing, species classification analysis and abundance analysis, functional annotation and pathway analysis, data standardization and dimensionality reduction processing and feature screening processing on the initial intestinal microorganism sequencing data, and screen out intestinal microorganism assessment-related features; a gene collection and processing submodule, which is used to upload gene initial detection data in multiple data formats provided by a third-party gene detection platform, and perform format conversion, field mapping processing and genetic feature extraction on the initial gene detection data, and screen out SNP sites significantly associated with obesity, wherein the SNP site significantly associated with obesity refers to a SNP site with a β value or OR value associated with obesity greater than the first type of preset value, and the polygenic risk score value is calculated through the individual polygenic risk score calculation formula; a lifestyle and diet data collection and processing submodule, which is used to input dietary habit information and lifestyle information, and obtain dietary habit scores and lifestyle scores through weighted integration processing.

[0030] For further information, see Attachment Figure 3The intestinal microorganism collection and processing submodule includes: an intestinal microorganism collection submodule, which is used to collect fecal samples of target users, extract microbial DNA from fecal samples, and obtain initial intestinal microorganism sequencing data through 16S rRNA sequencing or metagenomic sequencing; a quality control and sequence alignment submodule, which is used to check the quality of intestinal microorganism sequencing data, filter out low-quality sequences and connector sequences contained in the initial intestinal microorganism sequencing data, and obtain intestinal microorganism sequencing processing data, wherein the low-quality sequence refers to a sequence quality score less than the second category preset value; and perform sequence alignment processing on the intestinal microorganism sequencing processing data and the reference microorganism genome database to obtain intestinal microorganism sequencing alignment result data; a species classification and abundance analysis submodule, which is used to perform species classification on the intestinal microorganism sequencing alignment result data to obtain species information and abundance information of the intestinal microbial community; a functional annotation and pathway analysis submodule , which is used to perform functional prediction on the species and abundance information of intestinal microorganisms according to the gene function database, obtain the functional potential information of intestinal microorganisms, and extract the functional pathway characteristics related to obesity based on the functional potential information of intestinal microorganisms; the data standardization and dimensionality reduction submodule is used to standardize the abundance information and functional pathway characteristics of intestinal microorganisms to obtain standardized intestinal microorganism data, and perform dimensionality reduction on the standardized intestinal microorganism data to extract the effective features of intestinal microorganisms; the feature screening submodule is used to call the microbial database of the knowledge base module, and use the abundance, species and functional pathway characteristics of intestinal microorganisms as constraints to screen out intestinal microorganism evaluation-related features from the effective features of intestinal microorganisms.

[0031] Furthermore, by setting up a quality control and sequence alignment submodule, an inspection tool (such as FastQC, etc.) can be used to check the quality of the intestinal microbial sequencing data so as to filter out low-quality sequences and adapter sequences, wherein low-quality sequences refer to sequences with a quality score less than a preset value of the second category (such as 20); the intestinal microbial sequencing processing data is then subjected to sequence alignment processing with a reference microbial genome database to obtain intestinal microbial sequencing alignment result data, wherein the reference microbial genome database includes but is not limited to Greengenes, SILVA or NCBI RefSeq, etc.; at the same time, the purpose of sequence alignment processing is to perform species identification, relative abundance analysis, function prediction, etc. After sequence alignment processing, OTU (operational taxonomic unit) or ASV (amplicon sequence variant), classification information, abundance table and other data (i.e., intestinal microbial sequencing alignment result data) can be obtained, and the obtained intestinal microbial sequencing alignment result data is used in the species classification and abundance analysis submodule, and the intestinal microbial sequencing alignment result data is further optimized and standardized through this submodule to effectively improve the accuracy of the data.

[0032] Furthermore, by setting up species classification and abundance analysis submodules, classification tools (such as QIIME2 or MetaPhlAn) can be used to classify the intestinal microbial sequencing comparison result data. For example, the intestinal microbial sequencing comparison result data can be clustered by OTU / ASV and then classified to determine the microbial species and relative abundance in fecal samples, and then output a microbial relative abundance table at the species level or a higher classification level (such as genus or phylum).

[0033] Furthermore, by setting up functional annotation and pathway analysis submodules, functional prediction tools (such as PICRUSt or Humann) can be used to predict the functional potential of intestinal microorganisms based on known gene function databases (such as KEGG, MetaCyc); at the same time, based on the functional potential information of intestinal microorganisms, functional pathway characteristics related to obesity such as metabolism, energy balance, and inflammation can be extracted.

[0034] Furthermore, by setting up data standardization and dimensionality reduction submodules, the abundance information and functional pathway characteristics of intestinal microorganisms can be standardized by using centralization or standard normal distribution transformation technology to eliminate data scale differences, thereby obtaining standardized intestinal microorganism data, and the standardized intestinal microorganism data can be reduced in dimension by principal component analysis (PCA) or non-negative matrix decomposition (NMF) to extract effective intestinal microbial features.

[0035] Furthermore, by setting up a feature screening submodule and combining the microbial research literature related to obesity (i.e., the evidence information corresponding to the intestinal microbial information), the intestinal microbial evaluation-related features are screened out from the intestinal microbial effective features based on the three dimensions of intestinal microbial abundance, species and functional pathway characteristics. The intestinal microbial evaluation-related features, such as the abundance of certain short-chain fatty acid-producing bacteria, are screened out.

[0036] For further information, see Attachment Figure 4The gene collection and processing submodule includes: a gene data collection submodule, which is used to upload the initial gene detection data in various data formats provided by a third-party gene detection platform; a format conversion submodule, which is used to convert the format of the initial gene detection data in various data formats based on a rule-based parsing algorithm to obtain the gene detection data in JSON format; a field mapping processing submodule, which is used to construct a field mapping table, and based on the constructed field mapping table, perform field mapping on the gene detection data to obtain the gene detection mapping data; a genetic feature extraction submodule, which is used to screen the SNP sites contained in the gene detection mapping data by calling the gene database of the knowledge base module, and screen out the SNP sites that are significantly associated with obesity, wherein the SNP sites that are significantly associated with obesity refer to the SNP sites whose β value or OR value associated with obesity is greater than the first type preset value, and the polygenic risk score value is calculated by the individual polygenic risk score calculation formula.

[0037] Furthermore, by setting up a gene data collection submodule, users are allowed to upload initial gene detection data provided by a third-party gene detection platform, so as to realize the integration of multiple data sources about initial gene detection data, and support multiple file formats, such as VCF variation calling format, PLINK format, etc.

[0038] Furthermore, by setting a format conversion submodule for unified format conversion of multiple file formats, a rule-based parsing algorithm (such as Python's Pandas) can be used to automatically identify the file type and convert it into a unified JSON format, so as to obtain genetic testing data in JSON format.

[0039] Furthermore, by setting up a field mapping processing submodule and constructing a field mapping table, which contains predefined standard fields such as gene loci, SNP numbers, allele frequencies, etc., field names from different sources are aligned based on the field mapping table to eliminate differences between data and errors caused by the differences, thereby improving data accuracy and reliability.

[0040] Furthermore, by setting up a genetic feature extraction submodule, SNP site data is used to interpret obesity-related gene variations in order to identify SNP sites that are significantly associated with obesity, and the individual's polygenic risk score is calculated through the identified SNP sites, and the polygenic risk score is used as one of the obesity risk assessment features.

[0041] Furthermore, the specific implementation steps of genetic feature extraction include: 1. Quality control (QC) of genetic detection mapping data, specifically: (1) screening low sequencing quality sites (e.g., SNPs with sequencing depth less than 5 will be filtered out), (2) eliminating sites with minor allele frequency (MAF) less than 0.01 to reduce the impact of rare variants, (3) excluding sites that deviate from Hardy-Weinberg equilibrium (e.g., p value < 0.0001) to ensure the stability of population genetics; 2. Integrate public gene annotation databases and obesity-related literature (i.e., evidence information on the interaction between gene variation and microorganisms) to achieve a comprehensive interpretation of SNP sites, specifically: (1) Combine public gene annotation databases such as dbSNP, ClinVar, Ensembl, etc. to obtain the gene position of the site, the type of functional variation (such as non-synonymous mutation, missense mutation, etc.), the protein function that may be affected, and related disease annotations, (2) query GWAS Catalog database, including SNP sites significantly associated with obesity and its phenotypes (such as BMI, fat distribution, etc.) (e.g., GWAS significance level p value < ), and at the same time, by searching the literature, SNP sites related to obesity are collected to form a genome-specific obesity risk site library, which is the gene database; 3. Site screening, specifically: (1) by calling the gene database, the sites are sorted according to the β value or OR value associated with obesity by SNP sites, and the top 100 SNP sites are screened; (2) the above ranking results are manually reviewed to screen out high-value sites that are clearly reported in the literature to be significantly associated with obesity, for example, FTO gene (rs9939609): related to lipid metabolism Related to appetite regulation, PPARG gene (rs1801282): affects adipocyte differentiation and insulin sensitivity, MC4R gene (rs17782313): regulates energy balance and satiety; (3) From the high-value sites that have been clearly reported in the literature to be significantly associated with obesity, the top 30 related candidate SNP sites were screened for subsequent polygenic risk score calculation; 4. Based on the 30 screened SNP sites, the individual polygenic risk score calculation formula was used to calculate the individual's polygenic risk score value. The polygenic risk score calculation formula for the individual is: ; In the formula, PRS represents the polygenic risk score, Gj represents the number of allele copies of an individual at the j-th SNP site, Gj takes one of 0, 1, and 2, Rj represents the β value or OR value associated with the j-th SNP site and obesity, and m represents the number of SNP sites used for calculation after screening.

[0042] For further information, see Attachment Figure 5The lifestyle and diet data collection and processing submodule includes: a lifestyle and diet data input submodule, which is used to input dietary habit information and lifestyle information, wherein the dietary habit information includes the intake frequency and intake amount of sugary diet, the intake frequency and intake amount of high-fat diet and fat source information, and the lifestyle information includes exercise information and sleep information; an integration processing submodule, which is used to use weighted integration technology to perform weighted integration processing on the input dietary habit information and lifestyle information to generate dietary habit scores and lifestyle scores.

[0043] Furthermore, exercise information includes but is not limited to the user's exercise frequency, duration and intensity, sleep information includes but is not limited to sleep duration, sleep quality and nighttime interruptions, sugary diet refers to sugary drinks and sweets, and the frequency and amount of sugary drink and sweets intake can be input to evaluate the high-sugar diet type, high-fat diet refers to fried food, and the frequency, amount and fat source of fried food intake can be input to evaluate the high-fat diet type; by weighted integration of eating habit information and lifestyle information, eating habit scores and lifestyle scores can be generated, so that they can be used as one of the input features of a multi-classification obesity risk prediction model, and work together with biological characteristics such as genes and microorganisms to predict the user's personalized obesity risk.

[0044] In some embodiments, the risk prediction model module is used to input the multi-dimensional assessment preprocessing results into the trained multi-classification obesity risk prediction model, output the probability distribution of the target user belonging to high risk, medium risk, and low risk, and calculate the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtain the risk classification level result of the target user according to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field.

[0045] For further information, see Attachment Figure 6The risk prediction model module includes: a data processing and integration submodule, which is used to obtain the historical multi-dimensional evaluation preprocessing results and historical actual obesity risk values ​​of multiple users, standardize and feature encode the historical multi-dimensional evaluation preprocessing results, and obtain the historical multi-dimensional standardized feature vector, wherein the historical multi-dimensional standardized feature vector includes the standardized feature vector of intestinal microorganism data, the standardized feature vector of gene data, the standardized feature vector of dietary habit data, and the standardized feature vector of lifestyle data; a feature selection submodule, which is used to combine the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, and call the microbial database, gene database, dietary database, and lifestyle database of the knowledge base module to screen out the key features related to obesity risk; a model construction and training submodule, which is used to select the key features related to obesity risk according to the screened features. The key features related to obesity risk obtained are used to construct and train a multi-classification obesity risk prediction model based on the Softmax regression algorithm. The multi-classification obesity risk prediction model is used to calculate the probability distribution of each user belonging to high risk, medium risk, and low risk; the risk score calculation and classification submodule is used to input the multi-dimensional evaluation preprocessing results of the target user into the trained multi-classification obesity risk prediction model, output the probability distribution of the target user belonging to high risk, medium risk, and low risk, and calculate the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtain the risk classification level result to which the target user belongs according to the mapping relationship formed between the comprehensive obesity risk score and the risk classification level field; the model update and optimization submodule is used to continuously update and optimize the multi-classification obesity risk prediction model by using incremental learning technology.

[0046] Furthermore, the historical intestinal microbial assessment-related features were standardized and feature encoded to obtain the standardized feature vector of intestinal microbial data. The specific implementation process is as follows: (1) Data standardization: the z-score standardization calculation formula was used to standardize the intestinal microbial assessment-related features and convert them into a form with a mean of 0 and a standard deviation of 1. The z-score standardization calculation formula is: ; In the formula, x is the value of the intestinal microbial assessment-related feature, μ is the sample mean, and σ is the sample standard deviation; (2) Feature encoding, encoding the abundance information or functional pathway features of intestinal microorganisms into a numerical vector to form a standardized feature vector. For example, feature type: microbial species abundance, diversity index, functional feature value. The obtained intestinal microbial data standardized feature vector is: Xmicrobiome=[species A abundance, species B abundance, function 1 value, function 2 value].

[0047] Furthermore, the historical polygenic risk score values ​​are standardized and feature encoded to obtain the standardized feature vector of genetic data. The specific implementation process is as follows: (1) Data standardization, using a z-score standardization calculation formula similar to the above, the polygenic risk score value is subjected to z-score standardization conversion. In this case, x in the z-score standardization calculation formula is the polygenic risk score value (PRS). (2) Feature encoding, which can be done by directly inputting the standardized score or performing binning and then encoding it into a feature vector to obtain the standardized feature vector of genetic data.

[0048] Furthermore, the historical dietary habit scores and lifestyle scores are standardized and feature encoded to obtain the standardized feature vectors of dietary habit data and the standardized feature vectors of lifestyle data. The specific implementation process is as follows: (1) Data standardization: a z-score standardized calculation formula similar to the above is used to perform z-score standardized conversion on the quantified dietary habit scores and lifestyle scores. In this case, x in the z-score standardized calculation formula is the dietary habit score or the lifestyle score; (2) Feature encoding: the standardized score is directly encoded into a feature vector to obtain the standardized feature vectors of dietary habit data and the standardized feature vectors of lifestyle data. It should be noted that the quantification process of dietary habit score and lifestyle score is as follows: (1) assigning specific scores based on the collected dietary habit information and lifestyle information, for example, high-sugar diet: daily sugar intake, frequency of sweets; high-fat diet: fat intake (g / day), frequency of fried foods; exercise: collecting exercise frequency (times / week), exercise duration (minutes / time) and intensity; sleep: collecting daily sleep duration, sleep time, deep sleep ratio, etc., assigning preset specific scores to dietary habit information, i.e. the aforementioned high-sugar diet and high-fat diet, and then obtaining the high-sugar diet score and high-fat diet score. At the same time, for the lifestyle score, a lifestyle scoring system is constructed based on various indicators of exercise and sleep. For example, if the sleep duration is less than 4 hours or more than 10 hours, 10 points will be deducted, etc., to obtain the lifestyle score. (2) Score calculation: calculate the comprehensive score of high-sugar and high-fat diets according to predefined calculation rules. For example, pre-assign preset weights to high-sugar diet and high-fat diet, and then calculate the comprehensive score (i.e. dietary habit score) based on the obtained high-sugar diet score or high-fat diet score × the corresponding weight.

[0049] For further information, see Attachment Figure 7The feature selection submodule includes: a feature matrix construction submodule, which is used to construct a feature matrix based on the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, with each user as a row and each user's historical multi-dimensional standardized feature vector as a column; a target variable preparation submodule, which is used to take the historical actual obesity risk value of each user as the target variable; a Lasso regression modeling submodule, which is used to establish a regularized inverse Lasso regression objective function based on the feature matrix and the target variable, wherein the Lasso regression objective function is: ; In the formula, yi represents the user's actual historical obesity risk value, i represents the predicted obesity risk value, βi represents the regression coefficient of the historical multidimensional standardized feature vector i, λ represents the regularization parameter, which is used to control the strength of L1 regularization, and the regularization parameter λ is selected by cross-validation, and n represents the total number of users; a feature screening submodule is used to screen out features that are significantly correlated with obesity risk based on L1 regularization of the Lasso regression objective function, wherein the features that are significantly correlated with obesity risk refer to the features that are determined by the value of βi approaching zero when the L1 regularization of the Lasso regression objective function is performed, and the microbial database, gene database, diet database and lifestyle database of the knowledge base module are called to verify the biological significance of the features that are significantly correlated with obesity risk, and the features that have passed the biological significance verification are used as key features related to obesity risk.

[0050] Furthermore, in the process of constructing the feature matrix, based on the historical multidimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, the historical multidimensional standardized feature vectors are integrated into a feature matrix, in which each row represents a user, and each column represents the historical multidimensional standardized feature vector of the corresponding user (such as the abundance of a certain microbial species or the encoding of a certain gene locus).

[0051] Furthermore, a target variable preparation submodule is provided to set a target variable, wherein the target variable is the user's historical actual obesity risk value, and the historical actual obesity risk value can be obtained by classification of BMI, waist circumference index or body fat percentage.

[0052] Furthermore, by setting the Lasso regression modeling submodule, the Lasso regression method can be used to optimize the target by minimizing the following Lasso regression objective function: ; In the formula, yi represents the user's actual historical obesity risk value, i represents the predicted obesity risk value, βi represents the regression coefficient of the historical multidimensional standardized feature vector i, λ represents the regularization parameter, which is used to control the strength of L1 regularization, and the regularization parameter λ is selected by cross-validation to ensure the generalization ability of the model, and n represents the total number of users.

[0053] Furthermore, in the feature screening process, the L1 regularization based on the Lasso regression objective function will make the value of the weight βi approach zero, thereby automatically screening out features that are significantly related to obesity risk. Subsequently, by calling the microbial database, gene database, dietary database, and lifestyle database of the knowledge base module, the biological significance of the features that are significantly related to obesity risk is verified (such as whether a certain microbial species is related to short-chain fatty acid metabolism). The features that have passed the biological significance verification are used as key features related to obesity risk (including microbial species, gene mutation sites, dietary patterns, etc. that are significantly related to obesity risk). These key features are used for the subsequent training and verification of multi-classification obesity risk prediction models.

[0054] Furthermore, based on the screened key features related to obesity risk, the Softmax regression algorithm is used to construct an initial multi-classification obesity risk prediction model, which is used to calculate the probability distribution of each user belonging to high risk, medium risk, and low risk. The multi-classification obesity risk prediction model is: ; In the formula, P(y=c|X) represents the probability that user X is classified as risk level c; e represents the base of the natural logarithm, which is approximately equal to 2.718; Softmax uses an exponential function to ensure that the output value is positive and can amplify larger values, making the model more discriminative; Represents the intercept term (i.e., bias term) of category c, which is used to adjust the basic probability of the category; represents the j-th feature weight corresponding to category c, Represents the score of category c under given feature X; represents the exponential sum of all categories, which is used for normalization to ensure that the sum of the probabilities of all categories is 1; c represents the current category (such as low risk, medium risk or high risk); k represents the index used to traverse all categories (0, 1, and 2 represent low risk, medium risk, and high risk, respectively). The multi-classification obesity risk prediction model estimates parameters by maximizing the log-likelihood function. The formula for maximizing the log-likelihood function is: ; where n represents the total number of samples in the training data set (i.e., the number of users); i represents the i-th sample; y i The label representing each user's actual obesity risk category, usually represented by an integer (such as 0, 1, 2) or a One-Hot code; Indicates that given the input feature X i In the case of i The probability is calculated by Softmax.

[0055] Furthermore, during the model training process, a historical data set containing known obesity risks (i.e., multi-dimensional evaluation data) is used. The historical data set includes: the user's actual obesity risk value, intestinal microbial sequencing data, obesity-related gene detection data, dietary habit information and lifestyle information, etc., and the aforementioned data are preprocessed to obtain the historical multi-dimensional evaluation preprocessing results (i.e., intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores and lifestyle scores). The multi-classification obesity risk prediction model is trained and the historical data set is divided into a training set and a test set according to a preset ratio (e.g., 7:3). K-fold cross-validation (usually K=10) is used to ensure that the model can generalize to unseen data. During each training, the following model evaluation indicators are calculated, namely: (1) accuracy, which is used to calculate the percentage of correctly classified users; (2) recall, which is used to determine the model's ability to identify high-risk users; (3) F1-score, which is a trade-off indicator that comprehensively considers accuracy and recall; and (4) AUC (area under the curve), which is used to evaluate the classification ability of the model. These indicators are used to evaluate the performance of the model at different risk levels, and the optimal model is selected for deployment.

[0056] Furthermore, after the model training is completed, the multi-dimensional evaluation preprocessing results of the target user are input into the trained multi-classification obesity risk prediction model, and the probability distribution of the target user belonging to high risk, medium risk, and low risk is output to indicate the possibility of the target user belonging to high risk, medium risk, and low risk, and a weight is assigned to each risk level, such as the comprehensive risk score (Risk Score) calculation formula: ; According to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field, the risk classification level result to which the target user belongs is obtained, wherein the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field is, for example, low risk classification level: comprehensive obesity risk score <0.7, medium risk classification level: 0.7≤comprehensive obesity risk score <1.3, high risk classification level: comprehensive obesity risk score ≥1.3.

[0057] Furthermore, after the model is deployed and applied, as new data continues to be input, the model is regularly retrained using new data sets and the model parameters are fine-tuned. Incremental learning techniques can be used to continuously update the model parameters without completely retraining the entire model. By continuously optimizing the model, it is ensured that the accuracy of the model in predicting obesity risk can be gradually improved over time. At the same time, the model can also be tuned through user feedback and new key features can be added to improve prediction accuracy.

[0058] Furthermore, the specific implementation process of using incremental learning technology to continuously update and optimize the multi-classification obesity risk prediction model is as follows: (1) Regularly collect new data input by users, standardize it and convert it into a data format acceptable to the model; (2) Update the model weights. While locking the early layers, update the weights of the later layers of the model through small batch training to make the model adapt to the new data characteristics; (3) Online learning strategy, balance new and old data, dynamically adjust weights or use replay mechanism to ensure that the model takes into account the learning effects of both historical and new data; (4) Verification and optimization: Evaluate the model performance through verification indicators (such as AUC, F1-score), and fine-tune parameters or expand new data when necessary.

[0059] In some embodiments, the knowledge base module is used to construct a microbial database, a gene database, a dietary database, and a lifestyle database, and based on the multi-dimensional evaluation data of the target user, screen out multi-dimensional risk assessment features from the microbial database, the gene database, the dietary database, and the lifestyle database, and use a pattern matching algorithm to perform correlation analysis on the multi-dimensional risk assessment features and key features related to obesity risk to obtain correlation analysis results, and based on the correlation analysis results, screen out key risk impact features related to obesity according to evidence-based weights, and store corresponding health management recommendation information based on the risk classification level results.

[0060] For further information, see Attachment Figure 8 The knowledge base module includes: a microbial database for storing intestinal microbial information related to obesity and evidence information corresponding to the intestinal microbial information; a gene database for recording gene variations directly related to obesity and evidence information on the interaction between gene variations and microorganisms; a dietary database for recording information and evidence information on the impact of different dietary patterns on obesity, and information and evidence information on the regulatory effects of different dietary patterns on intestinal microorganisms; a lifestyle database for recording information and evidence information on the impact of different lifestyles on obesity, and information and evidence information on the regulatory effects of different lifestyles on intestinal microorganisms; an association analysis submodule for screening out multidimensional risk assessment features from the microbial database, gene database, diet database, and lifestyle database based on the multidimensional evaluation data of the target user, and using a pattern matching algorithm to perform correlation analysis on the multidimensional risk assessment features and key features related to obesity risk, and obtain correlation analysis results, and based on the correlation analysis results, screen out key risk impact features related to obesity according to the evidence-based weights; a health management recommendation storage submodule for storing corresponding health management recommendation information based on the risk classification level results.

[0061] Furthermore, the microbial database is mainly used to store intestinal microbial information related to obesity and evidence information corresponding to the intestinal microbial information, such as microbial ID, microbial name (such as Bacteroides fragilis), functional classification (such as short-chain fatty acid production), correlation with obesity risk, metabolic function description: briefly describe the metabolic effect of the microorganism and its potential impact on host health (such as producing butyrate, which helps regulate intestinal inflammation), literature (such as: [1] Smith et al. (2020), "Role of Gut Microbiota in Obesity") and other fields.

[0062] Furthermore, the gene database is mainly used to record genetic variants directly related to obesity and evidence of the interaction between genetic variants and microorganisms, such as gene ID, SNP site (such as: specific single nucleotide polymorphism site, variant rs9939609 associated with obesity risk), functional description: describing the metabolic function of the gene and whether it is associated with obesity risk (such as: affecting energy metabolism), β value or OR value, microbial ID (associated microbial ID), interaction type (nature of the interaction between the gene and the microorganism (such as: promotion or inhibition), interaction description: detailed description of the biological mechanism of the interaction (such as: this SNP variant leads to insulin resistance, thereby promoting the increase in the abundance of certain microorganisms), literature (such as: [2] Zhang et al. (2022), "Gene-Microbiota Interactions"). It should be noted that the SNP sites related to obesity are mainly distributed in pathways such as metabolic regulation, neural signaling, fat distribution and development, inflammation and immunity. The FTO gene (SNP site: rs9939609) in the metabolic pathway regulates appetite and fat metabolism, and is the core of obesity research; the BDNF gene (SNP site: rs6265) in the neural signaling pathway indirectly affects appetite by regulating neural development; the LYPLAL1 (SNP site: rs2605100) gene related to fat distribution is related to abdominal fat deposition; the IL6 gene (SNP site: rs1800795) in the inflammatory pathway affects obesity and chronic inflammation by regulating the inflammatory factor IL-6; in gene-environment interactions, the TCF7L2 gene (SNP site: rs7903146) significantly affects the risk of obesity and type 2 diabetes. In addition, GWAS studies have also found that genes such as TMEM18 and SH2B1 are closely related to obesity. Although their specific functions have yet to be further clarified, the synergistic effects of these genes will still comprehensively affect the individual's obesity risk.

[0063] Furthermore, the dietary database is mainly used to record information and evidence on the impact of different dietary patterns on obesity, and information and evidence on the regulatory effects of different dietary patterns on intestinal microorganisms, for example, dietary type: description of the specific dietary pattern (e.g., high-sugar diet), impact description: the association between this dietary pattern and obesity (e.g., high-sugar diet increases fat storage), microbial ID (microbial ID associated with this dietary pattern), impact direction: the impact of diet on microorganisms (e.g., increase or decrease), impact mechanism description: detailed description of the biological mechanism of how this diet affects the microbial community (e.g., high-sugar diet promotes the proliferation of specific yeasts, thereby inhibiting the growth of probiotics), literature; citations of research supporting the interaction between this diet and microorganisms (e.g., [3] Johnson et al. (2021), "Dietary Influences on Gut Microbiota"), etc.

[0064] Furthermore, the lifestyle database is used to record information and evidence on the impact of different lifestyles on obesity, and information and evidence on the regulatory effects of different lifestyles on intestinal microorganisms, for example, lifestyle variables: description of specific lifestyle factors (such as exercise frequency, sleep quality), impact description (such as: insufficient exercise frequency increases the risk of obesity), microbial ID, impact description: the specific impact of lifestyle on the microbial community (such as: an increase in weekly exercise frequency will lead to an increase in Lactobacillus abundance), impact direction: the nature of the impact (such as: positive or negative), literature (such as: [4] Lee et al. (2023), "Impact of Lifestyle on Gut Microbiota"), etc.

[0065] Furthermore, the knowledge base module involved in the present application may also include a risk prediction model database, which is mainly used to store the model's training data, prediction results and model parameters to facilitate model updates and performance evaluation; for example, model ID (a unique identifier that identifies different versions of prediction models), training data source (describing the source and size of the data set used for model training), parameter weights, model performance indicators (recording the model's prediction accuracy, sensitivity, specificity and other performance indicators), update history (recording the model's update log, including the date, content and reason for the update), etc.

[0066] Furthermore, by setting up an association analysis submodule, based on the multi-dimensional evaluation data of the target user, the multi-dimensional evaluation data of the target user is compared with the obesity-related information (such as specific microbial species, gene mutations, dietary patterns, etc.) recorded in the microbial database, gene database, diet database and lifestyle database, so as to identify the corresponding multi-dimensional risk assessment features, and then a pattern matching algorithm (such as support vector machine, random forest, etc.) is used to analyze the correlation between the user's multi-dimensional risk assessment features and the key features related to obesity risk, so as to determine whether the user's multi-dimensional risk assessment features are associated with the key features, and obtain the corresponding correlation analysis results, and based on the correlation analysis results, according to the evidence-based weights (such as the research evidence level and statistical significance recorded in the knowledge base), the core factors affecting obesity risk (i.e., the key risk impact features related to obesity) are screened out as the basis for report generation. In addition, the model can also be continuously optimized according to the data in the knowledge base module, and the weights of each feature can be dynamically adjusted to improve the prediction accuracy of the model.

[0067] Furthermore, by setting up a health management advice storage submodule, the health management advice storage submodule can store corresponding health management advice information according to the risk classification level results, such as high risk, medium risk, and low risk. The health management advice information is summarized and summarized based on the relevant information found. Among them, the health management advice information mainly comes from authoritative medical guidelines (such as "AHA New Guidelines on Obesity and Cardiovascular Risk", "AACE / ACE Guidelines: Comprehensive Medical Management of Obese Patients", "China Obesity Diagnosis and Treatment Guidelines", "2019 European Practice Guidelines: Management of Obesity in Adults in Primary Care", etc.), academic research (systematic reviews, clinical trials, etc.), clinical practice experience (Obesity Management Center, Expert Consensus), and health behavior intervention recommendations (marketing nutrition, exercise, psychological intervention, etc.); the specific summary and induction process is as follows: first, build a knowledge extraction framework, and set core management dimensions around different risk levels (high risk, medium risk, low risk), including diet adjustment, exercise intervention, metabolic monitoring, psychological behavioral intervention, etc.; then, through information screening and classification and induction, extract scientific and feasible management suggestions from a large amount of literature and practice data (i.e. the aforementioned authoritative medical guidelines), and perform deduplication and optimization processing to obtain optimized health management suggestion information to ensure the accuracy and operability of the content; and store the optimized health management suggestion information in the health management suggestion storage submodule. When the report is generated, the system automatically matches the corresponding suggestions according to the risk classification level results of the target user, and provides users with accurate and personalized health management guidance.

[0068] In some embodiments, a report generation module is used to generate a personalized obesity risk assessment report based on the target user's comprehensive obesity risk score, the target user's risk classification level results, key risk impact characteristics related to obesity, evidence-based medical information corresponding to the key risk impact characteristics related to obesity, and health management recommendation information corresponding to the target user's risk classification level results.

[0069] Furthermore, the personalized obesity risk assessment report may include the target user's comprehensive obesity risk score, key risk impact characteristics related to obesity (such as specific microbial populations, gene mutations, dietary habits), the risk classification level results to which the target user belongs, and evidence-based medical information corresponding to the key risk impact characteristics related to obesity (such as evidence information and literature basis for the association or interaction of each key risk impact characteristic) to increase the user's trust in the report. It may also include personalized health management recommendations. At the same time, the personalized obesity risk assessment report will be presented in an easy-to-understand format and provide specific health management recommendations, such as adjusting the dietary structure, increasing the frequency of exercise, and improving lifestyle.

[0070] Further, assume that a user A provides the following data: (1) Microbiome data: Through 16S sequencing, the abundance of certain bacterial communities in the intestinal microorganisms of user A is significantly higher than the normal level; (2) Genetic testing data: The genetic testing results show that user A carries SNP sites related to obesity (such as rs9939609); (3) Dietary habit information: User A's daily diet tends to be high in sugar and high in fat; (4) Lifestyle information: User A has a low exercise frequency, short sleep time, and high stress. In this embodiment, the system evaluates the obesity risk of user A through the following steps: (1) Input and preprocess user A's microintestinal microbial sequencing data, genetic testing data, dietary habit information, and lifestyle information to obtain a multi-dimensional evaluation preprocessing result; (2) Input the multi-dimensional evaluation preprocessing result into the multi-classification obesity risk prediction model to obtain the probability distribution of the target user belonging to high risk, medium risk, and low risk, calculate the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtain the risk classification level result to which the target user belongs based on the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field. ; (3) Retrieve multidimensional risk assessment features related to user A’s intestinal microbial composition, SNP sites, dietary habits and lifestyle from the knowledge base. For example, high abundance of Bacteroides is highly correlated with obesity. A’s rs9939609 gene variant further increases the risk of obesity, while a high-sugar diet and low physical activity exacerbate this obesity risk; (4) Generate a personalized obesity risk assessment report, which points out that user A’s main obesity risk comes from specific microbial abundance, gene variants and unhealthy dietary habits, and recommends that user A reduce high-sugar diet, increase dietary fiber intake, and exercise regularly.

[0071] Embodiment 2

[0072] This embodiment provides a method for generating an obesity risk prediction report based on a knowledge base on the basis of the above embodiment 1. Fig. 9 , a knowledge base-based obesity risk prediction report generation system applied to Example 1, the method includes the following steps.

[0073] S1. Collect and preprocess the multi-dimensional evaluation data of the target user through the data collection module to obtain a multi-dimensional evaluation preprocessing result, wherein the multi-dimensional evaluation data includes initial intestinal microbial sequencing data, initial gene detection data, dietary habit information and lifestyle information; the multi-dimensional evaluation preprocessing result includes intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores and lifestyle scores.

[0074] S2. The multi-dimensional assessment preprocessing results are input into the trained multi-classification obesity risk prediction model through the risk prediction model module, and the probability distribution of the target user belonging to high risk, medium risk and low risk is output. The comprehensive obesity risk score of the target user is calculated based on the comprehensive risk score calculation formula. According to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field, the risk classification level result of the target user is obtained.

[0075] S3. Build a microbial database, a gene database, a diet database and a lifestyle database through the knowledge base module. Based on the multi-dimensional evaluation data of the target users, screen out multi-dimensional risk assessment features from the microbial database, the gene database, the diet database and the lifestyle database. Use a pattern matching algorithm to perform correlation analysis on the multi-dimensional risk assessment features and the key features related to obesity risk to obtain the correlation analysis results. Based on the correlation analysis results and according to the evidence-based weights, screen out the key risk impact features related to obesity. Based on the risk classification level results, store the corresponding health management recommendation information.

[0076] S4. Generate a personalized obesity risk assessment report through the report generation module based on the comprehensive obesity risk score of the target user, the risk classification level of the target user, the key risk impact characteristics related to obesity, the evidence-based medical information corresponding to the key risk impact characteristics related to obesity, and the health management recommendation information corresponding to the risk classification level of the target user.

[0077] Obviously, it should be understood by those skilled in the art that the above-mentioned various steps of the present invention can be performed in a manner different from the present invention, and the simulation method and experimental equipment include but are not limited to the above description. The above-mentioned various steps of the present invention can be performed in an order different from that here in some cases, and the steps shown or described above can be performed separately. Therefore, the present invention is not limited to any specific combination of hardware and software.

[0078] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. A knowledge base-based obesity risk prediction report generation system, characterized in that: include: A data collection module is used to collect and preprocess the multi-dimensional evaluation data of the target user to obtain a multi-dimensional evaluation preprocessing result, wherein the multi-dimensional evaluation data includes initial intestinal microbial sequencing data, initial gene detection data, dietary habit information and lifestyle information; the multi-dimensional evaluation preprocessing result includes intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores and lifestyle scores; The risk prediction model module is used to input the multi-dimensional assessment preprocessing results into the trained multi-classification obesity risk prediction model, output the probability distribution of the target user belonging to high risk, medium risk, and low risk, and calculate the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtain the risk classification level result of the target user according to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field; The knowledge base module is used to construct a microbial database, a gene database, a diet database and a lifestyle database. Based on the multi-dimensional evaluation data of the target user, the multi-dimensional risk assessment features are screened out from the microbial database, the gene database, the diet database and the lifestyle database. The pattern matching algorithm is used to perform a correlation analysis on the multi-dimensional risk assessment features and the key features related to the risk of obesity to obtain the correlation analysis results. Based on the correlation analysis results, the key risk impact features related to obesity are screened out according to the evidence-based weights, and the corresponding health management advice information is stored based on the risk classification level results. The report generation module is used to generate a personalized obesity risk assessment report based on the comprehensive obesity risk score of the target user, the risk classification level result of the target user, the key risk impact characteristics related to obesity, the evidence-based medical information corresponding to the key risk impact characteristics related to obesity, and the health management recommendation information corresponding to the risk classification level result of the target user.

2. A knowledge base-based obesity risk prediction report generation system, characterized in that: The data acquisition module comprises: The intestinal microorganism collection and processing submodule is used to collect fecal samples from target users, extract microbial DNA from fecal samples, obtain initial intestinal microorganism sequencing data through 16S rRNA sequencing or metagenomic sequencing, and perform quality control and sequence alignment processing, species classification analysis and abundance analysis, functional annotation and pathway analysis, data standardization and dimensionality reduction processing, and feature screening processing on the initial intestinal microorganism sequencing data to screen out intestinal microorganism assessment-related features; The gene collection and processing submodule is used to upload the initial gene detection data in various data formats provided by the third-party gene detection platform, and perform format conversion, field mapping processing and genetic feature extraction on the initial gene detection data, and screen out SNP sites significantly associated with obesity, wherein the SNP sites significantly associated with obesity refer to the SNP sites with a β value or OR value associated with obesity greater than the first type of preset value, and calculate the polygenic risk score value through the individual polygenic risk score calculation formula; The lifestyle and diet data collection and processing submodule is used to input dietary habit information and lifestyle information, and obtain dietary habit scores and lifestyle scores through weighted integration processing.

3. The obesity risk prediction report generation system based on a knowledge base according to claim 2, characterized in that: The intestinal microorganism collection and processing submodule includes: The intestinal microorganism collection submodule is used to collect fecal samples from target users, extract microbial DNA from the fecal samples, and obtain initial intestinal microbial sequencing data through 16S rRNA sequencing or metagenomic sequencing; The quality control and sequence alignment submodule is used to check the quality of the intestinal microorganism sequencing data, filter out the low-quality sequences and adapter sequences contained in the initial intestinal microorganism sequencing data, and obtain the intestinal microorganism sequencing processing data, wherein the low-quality sequence refers to a sequence quality score less than the second category preset value; and perform sequence alignment processing on the intestinal microorganism sequencing processing data and the reference microorganism genome database to obtain the intestinal microorganism sequencing alignment result data; The species classification and abundance analysis submodule is used to classify the intestinal microbial sequencing comparison result data and obtain the species and abundance information of the intestinal microbial community; The functional annotation and pathway analysis submodule is used to predict the function of intestinal microorganisms based on the gene function database, the species information and abundance information, obtain the functional potential information of intestinal microorganisms, and extract the functional pathway characteristics related to obesity based on the functional potential information of intestinal microorganisms; The data standardization and dimensionality reduction submodule is used to standardize the abundance information and functional pathway characteristics of intestinal microorganisms to obtain standardized intestinal microorganism data, and to reduce the dimensionality of the standardized intestinal microorganism data to extract effective intestinal microorganism features; The feature screening submodule is used to call the microbial database of the knowledge base module, and screen out intestinal microbial assessment-related features from the intestinal microbial effective features based on the abundance, type and functional pathway characteristics of intestinal microorganisms as constraints.

4. The obesity risk prediction report generation system based on a knowledge base according to claim 2, characterized in that: The gene collection and processing submodule includes: Gene data collection submodule, used to upload initial gene detection data in various data formats provided by third-party gene detection platforms; The format conversion submodule is used for converting the format of the initial gene detection data in various data formats based on the rule-based parsing algorithm to obtain the gene detection data in JSON format; The field mapping processing submodule is used to construct a field mapping table, and based on the constructed field mapping table, perform field mapping on the gene detection data to obtain gene detection mapping data; The genetic feature extraction submodule is used to screen the SNP sites contained in the gene detection mapping data by calling the gene database of the knowledge base module, and screen out the SNP sites that are significantly associated with obesity, wherein the SNP sites that are significantly associated with obesity refer to the SNP sites whose β value or OR value associated with obesity is greater than the first type of preset value, and calculate the polygenic risk score value through the individual polygenic risk score calculation formula.

5. The obesity risk prediction report generation system based on a knowledge base according to claim 4, characterized in that: The polygenic risk score calculation formula for the individual is: In the formula, PRS represents the polygenic risk score value, Gj represents the number of allele copies of an individual at the j-th SNP site, Gj takes one of 0, 1, and 2, Rj represents the β value or OR value associated with the j-th SNP site and obesity, and m represents the number of SNP sites used for calculation after screening.

6. The obesity risk prediction report generation system based on a knowledge base according to claim 2, characterized in that: The lifestyle and diet data collection and processing submodule includes: The lifestyle and diet data input submodule is used to input diet habit information and lifestyle information, wherein the diet habit information includes the intake frequency and amount of sugary diet, the intake frequency and amount of high-fat diet and fat source information, and the lifestyle information includes exercise information and sleep information; The integration processing submodule is used to perform weighted integration processing on the input eating habit information and lifestyle information using weighted integration technology to generate eating habit scores and lifestyle scores.

7. The obesity risk prediction report generation system based on a knowledge base according to claim 1, characterized in that: The risk prediction model module includes: A data processing and integration submodule is used to obtain historical multi-dimensional assessment preprocessing results and historical actual obesity risk values ​​of multiple users, standardize and feature encode the historical multi-dimensional assessment preprocessing results, and obtain historical multi-dimensional standardized feature vectors, wherein the historical multi-dimensional standardized feature vectors include intestinal microbial data standardized feature vectors, gene data standardized feature vectors, dietary habit data standardized feature vectors, and lifestyle data standardized feature vectors; The feature selection submodule is used to combine the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, as well as the microbial database, gene database, diet database and lifestyle database of the knowledge base module to screen out key features related to obesity risk; The model building and training submodule is used to build and train a multi-classification obesity risk prediction model based on the Softmax regression algorithm according to the key features related to obesity risk screened out. The multi-classification obesity risk prediction model is used to calculate the probability distribution of each user belonging to high risk, medium risk, and low risk; The risk score calculation and classification submodule is used to input the multi-dimensional assessment preprocessing results of the target user into the trained multi-classification obesity risk prediction model, output the probability distribution of the target user belonging to high risk, medium risk, and low risk, and calculate the comprehensive obesity risk score of the target user based on the comprehensive risk score calculation formula, and obtain the risk classification level result of the target user according to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field; The model update and optimization submodule is used to continuously update and optimize the multi-classification obesity risk prediction model using incremental learning technology.

8. The obesity risk prediction report generation system based on a knowledge base according to claim 7, characterized in that: The feature selection submodule comprises: A feature matrix construction submodule is used to construct a feature matrix based on the historical multi-dimensional standardized feature vectors and historical actual obesity risk values ​​of multiple users, with each user as a row and each user's historical multi-dimensional standardized feature vector as a column; A target variable preparation submodule is used to take each user's historical actual obesity risk value as the target variable; The Lasso regression modeling submodule is used to establish a regularized inverse Lasso regression objective function based on the feature matrix and the target variable, wherein the Lasso regression objective function is: In the formula, yi represents the user's actual historical obesity risk value, i represents the predicted obesity risk value, βi represents the regression coefficient of the historical multi-dimensional standardized feature vector i, λ represents the regularization parameter, which is used to control the strength of L1 regularization, and the regularization parameter λ is selected by cross-validation, and n represents the total number of users; The feature screening submodule is used to screen out features that are significantly correlated with obesity risk based on L1 regularization of the Lasso regression objective function, wherein the features that are significantly correlated with obesity risk are determined by the value of βi approaching zero when the Lasso regression objective function is L1 regularized, and the microbial database, gene database, diet database and lifestyle database of the knowledge base module are called to verify the biological significance of the features that are significantly correlated with obesity risk, and the features that have passed the biological significance verification are used as key features related to obesity risk.

9. The obesity risk prediction report generation system based on a knowledge base according to claim 1, characterized in that: The knowledge base module includes: A microbial database, used to store intestinal microbial information related to obesity and evidence information corresponding to the intestinal microbial information; Gene databases to record genetic variants directly associated with obesity and evidence of interactions between genetic variants and microbes; Diet database, used to record information and evidence on the effects of different dietary patterns on obesity, and the regulatory effects of different dietary patterns on intestinal microorganisms; Lifestyle database, used to record information and evidence on the impact of different lifestyles on obesity, and information and evidence on the regulatory effects of different lifestyles on intestinal microorganisms; The association analysis submodule is used to screen out multidimensional risk assessment features from the microbial database, gene database, diet database and lifestyle database based on the multidimensional assessment data of the target user, and use a pattern matching algorithm to perform a correlation analysis on the multidimensional risk assessment features and key features related to obesity risk to obtain the correlation analysis results. Based on the correlation analysis results and according to the evidence-based weights, the key risk impact features related to obesity are screened out; The health management advice storage submodule is used to store corresponding health management advice information based on the risk classification level results.

10. A method for generating an obesity risk prediction report based on a knowledge base, characterized in that: include: The multi-dimensional evaluation data of the target user is collected and pre-processed by the data collection module to obtain a multi-dimensional evaluation pre-processing result, wherein the multi-dimensional evaluation data includes initial intestinal microbial sequencing data, initial gene detection data, dietary habit information and lifestyle information; the multi-dimensional evaluation pre-processing result includes intestinal microbial evaluation association characteristics, polygenic risk score values, dietary habit scores and lifestyle scores; The multi-dimensional assessment preprocessing results are input into the trained multi-classification obesity risk prediction model through the risk prediction model module, and the probability distribution of the target user belonging to high risk, medium risk and low risk is output. The comprehensive obesity risk score of the target user is calculated based on the comprehensive risk score calculation formula, and the risk classification level result of the target user is obtained according to the mapping relationship formed by the comprehensive obesity risk score and the risk classification level field; The microbial database, gene database, diet database and lifestyle database are constructed through the knowledge base module. Based on the multi-dimensional evaluation data of the target users, multi-dimensional risk assessment features are screened out from the microbial database, gene database, diet database and lifestyle database. The pattern matching algorithm is used to perform correlation analysis on the multi-dimensional risk assessment features and the key features related to obesity risk to obtain the correlation analysis results. Based on the correlation analysis results and according to the evidence-based weights, the key risk impact features related to obesity are screened out, and based on the risk classification level results, the corresponding health management recommendation information is stored. The report generation module generates a personalized obesity risk assessment report based on the target user's comprehensive obesity risk score, the target user's risk classification level, key risk impact characteristics related to obesity, evidence-based medical information corresponding to the key risk impact characteristics related to obesity, and health management recommendation information corresponding to the target user's risk classification level.

Images