Method for screening intestinal microorganism markers of chronic diseases and constructing early screening model

Abstract

The application discloses a health and chronic disease intestinal microorganism marker screening method, comprising the following steps: 1) obtaining sample intestinal microorganism metagenome relative abundance data and clinical information; 2) screening samples covering health and chronic disease phenotypes; 3) normalizing the intestinal microorganism metagenome relative abundance data of each sample; 4) in the training set, using nested cross-validation and sequence feature selection methods, screening phenotype-specific intestinal microorganism markers for health and each type of chronic disease phenotype respectively. The application also discloses a method for constructing a chronic disease early screening model based on the combination of the microorganism markers screened by the above method. The application realizes early precise screening and risk assessment of 11 common chronic diseases by screening intestinal microorganism markers and establishing a chronic disease early screening model, and can be used for distinguishing between healthy people and chronic disease risk people, and accurately identifying high-risk individuals of various chronic diseases.

Description

Technical Field

[0001] This invention relates to the field of biology, particularly to risk prediction in the early stages of chronic diseases, and more specifically, to the screening of gut microbiota biomarkers for chronic diseases, and a method for constructing early screening models for chronic diseases based on the screened gut microbiota biomarkers. Background Technology

[0002] Chronic diseases (CMDs) are a class of diseases with long courses and slow progression. They have become a leading cause of death and disability worldwide, posing a significant impact on socioeconomic development and individual health. CMDs not only seriously endanger individual health but also place a heavy burden on national medical resources, social productivity, and economic development. Common chronic diseases include, but are not limited to, metabolic diseases (such as type 2 diabetes and impaired glucose tolerance), blood pressure abnormalities, cardiovascular diseases, digestive system diseases (such as inflammatory bowel disease, colonic diseases, fatty liver, and cirrhosis), mental illnesses (such as schizophrenia), tumors (such as melanoma), and autoimmune diseases (such as ankylosing spondylitis).

[0003] Chronic diseases are generally characterized by long incubation periods, atypical early symptoms, and significant individual variability, making early clinical identification extremely difficult. Many patients are diagnosed at middle or late stages, missing the optimal intervention window. Early screening, a crucial component of chronic disease prevention and control, can significantly improve early diagnosis and treatment rates, reduce mortality and disability rates, while also alleviating the burden on healthcare, improving patient prognosis, and enhancing quality of life. Therefore, clinical practice and health management heavily rely on sensitive, accurate, and scalable screening methods. Currently, physical examination centers, community clinics, specialist follow-up clinics, and personalized health management institutions have become important venues for early screening and risk prediction of chronic diseases.

[0004] However, existing chronic disease screening methods have significant limitations. Questionnaires and physical examinations are insufficient for identifying early, occult lesions; while imaging examinations (such as CT and MRI) can provide structural information, they are expensive and pose radiation or invasive risks, making them unsuitable for frequent or large-scale screening; hematological or single biomarker tests can provide some information, but are mostly limited to single-disease identification and cannot cover the needs of screening for multiple diseases simultaneously. Furthermore, many chronic diseases share certain similarities in clinical presentation, easily leading to misdiagnosis and missed diagnosis. Therefore, there is an urgent clinical need for a new, non-invasive, sensitive screening technology capable of simultaneous analysis of multiple diseases, particularly suitable for follow-up monitoring of healthy individuals undergoing physical examinations, sub-healthy individuals, and high-risk groups.

[0005] In recent years, the gut microbiota, as an important regulator of host health, has gradually become a new focus in chronic disease research and clinical practice. Numerous studies have shown that the gut microbiota of patients with different chronic diseases exhibits characteristic abnormalities in abundance profiles, with changes in some microbiota closely related to disease risk, subtyping, and prognosis. With the development of metagenomic sequencing and multi-omics technologies, the gut microbiota is gradually demonstrating unique advantages in disease diagnosis, risk prediction, and efficacy assessment. Compared to traditional methods, it is a non-invasive, stable, and multi-disease-supporting source of bioinformatics, making it particularly suitable for clinical scenarios such as health checkups, personalized medicine, chronic disease follow-up monitoring, and public health population screening. However, there is still a lack of systematic, widely applicable combinations of representative key microbial biomarkers and standardized screening methods for various common chronic diseases. Summary of the Invention

[0006] One of the technical problems to be solved by the present invention is to provide a method for screening gut microbiota biomarkers for health and chronic diseases.

[0007] To address the aforementioned technical problems, the present invention provides a method for screening gut microbiota biomarkers for healthy individuals and those with chronic diseases, comprising the following steps: 1) Obtain relative abundance data of gut microbiota metagenomics and clinical information from the samples; 2) Screening samples that cover both healthy and chronic disease phenotypes; 3) Randomly divide the samples selected in step 2) into training set and test set, and normalize the relative abundance data of gut microbial metagenomics for each sample; 4) In the training set, nested cross-validation and sequence feature selection methods were used to screen phenotype-specific gut microbiota biomarkers for healthy phenotypes and each type of chronic disease phenotype.

[0008] The chronic diseases include one or more of the following: type 2 diabetes, inflammatory bowel disease, schizophrenia, colitis, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver.

[0009] Step 2), the preferred screening criteria include: using Illumina high-throughput sequencing, a cohort sample size of no less than 50, samples from non-infectious and acute diseases, and samples from feces of adults with no record of antibiotic use and a body mass index between 18.5 and 25.

[0010] Step 3), the sample size ratio of the training set and the test set is preferably 8.5:1.5; the normalization process preferably adopts the central logarithmic transformation method.

[0011] Step 4), the preferred screening method includes the following steps: Randomly divide the training set sample data into a biomarker discovery set and a biomarker evaluation set; in the biomarker discovery set, select one type of phenotype samples as target samples and the remaining samples as control samples. Randomly select some samples from the control samples to form a random biomarker discovery subset for that phenotype, together with the target samples. For each random biomarker discovery subset, use nested cross-validation to screen candidate microbial features that can specifically distinguish the corresponding phenotype and their phenotype specificity scores. Repeat the above process multiple times, calculate the average phenotype specificity score of each microorganism in all repetitions and rank them. Then, use a sequence forward feature selection method (preferably, gradually stacking microbial features according to phenotype specificity from high to low) to construct candidate microbial combinations of different sizes. Train a random forest model to predict the probability of each phenotype for each sample. When the predicted probability given by the candidate microbial combination for the target phenotype sample is significantly higher than that for non-target phenotype samples, determine that candidate microbial combination as the candidate microbial biomarker combination for the target phenotype; use the biomarker evaluation set to independently evaluate each candidate microbial biomarker combination, and finally obtain the gut microbial biomarker combination corresponding to each phenotype. The nested cross-validation method uses three-fold cross-validation in the inner layer to determine the optimal regularization parameters for the LASSO regression model. Under the optimal regularization parameters, the outer layer constructs and trains the LASSO regression model. The regression coefficients for each microorganism in the LASSO regression model represent the phenotypic specificity score of each microorganism's ability to discriminate the target phenotype. The method of independently evaluating each candidate microbial biomarker combination using a biomarker evaluation set further includes the following steps: based on the candidate microbial biomarker combinations, a random forest method is used to distinguish between samples of the target phenotype and non-target phenotypes, and the predicted probability of each sample belonging to the target phenotype is output; the F1 score is calculated by combining the predicted probability and the true phenotype label, and the candidate microbial biomarker combination with the highest F1 score is selected as the final gut microbial biomarker combination for that phenotype.

[0012] The second technical problem this invention aims to solve is to provide a combination of gut microbiota biomarkers for healthy individuals and 11 chronic diseases, screened according to the above method. Wherein: The healthy phenotype gut microbial marker panel includes the following 51 gut microbial species: Faecalibacterium prausnitzii, Streptococcus thermophilus, Bifidobacterium angulatum, Finegoldia magna, Pseudoflavonifractor capillosus, Coprobacter sp., Denitrobacterium detoxificans, Prevotella sp. CAG 1124, Asaccharobactercelatus, Peptoniphilus lacrimalis, Anaerobiospirillum thomasii, Collinsellamassiliensis, Prevotella timonensis, Enorma massiliensis, Streptococcus infantarius, Butyricimonas synergistica, Ruminococcus sp. CAG 563, Bacteroides sp. D2, Massiliomicrobiota timonensis, Bacteroides uniformis, Prevotella buccalis, Bacteroides pectinophilus, Bifidobacterium catenulatum, Bilophilawadsworthia, Sharpea azabuensis, Porphyromonas sp. HMSC065F10, Bacteroidesstercorirosoris, Cutibacterium acnes, Mitsuokella jalaludinii, Lachnoclostridiumsp. An118, Leuconostoc gelidum, Blautia sp. CAG 257, Olsenella scatoligenes, Intestinimonas butyriciproducens, Anaerotruncus sp. CAG 528, Citrobacterfreundii, Actinomyces sp.oral taxon 448, Aeriscardovia aeriphila, Comamonaskerstersii, Alistipes shahii, Prevotella sp. CAG 520, Romboutsia ilealis, Clostridium clostridioforme, Megasphaera elsdenii, Escherichia marmotae, Victivallis vadensis, Catenibacterium mitsuokai, Lactococcus petauri, Clostridium sp. CAG 510, Megamonas hypermegale, Lactococcus garvieae;. The gut microbial marker combination for type 2 diabetes phenotype includes the following 25 gut microbial species: Collinsella massiliensis, Aeriscardovia aeriphila, Gardnerella vaginalis, Eubacterium sp. CAG 251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemaniafiliformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaera elsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacterium varium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonas virosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans; The intestinal microbial biomarker set for inflammatory bowel disease phenotypes includes the following 27 intestinal microbial species: Enterorhabdus caecimuris, Enorma massiliensis, Bacteroides sp. CAG 144, Eisenbergiella massiliensis, Streptococcus lutetiensis, Anaeromassilibacillus sp. An250, Bacteroides vulgatus, Dorea longicatena, Lactobacillus sakei, Peptostreptococcus stomatis, Bacteroides sp. 43 108, Prevotella sp. 885, Fusobacterium nucleatum, Enterococcus faecium, Eubacterium sp. CAG 786, Blautia hansenii, Faecalibacterium prausnitzii, Blautia hydrogenotrophica, Lawsonibacterasaccharolyticus, and Dialister. pneumosintes, Oscillibacter sp. 57 20, Streptococcusmutans, Blautia obeum, Blautia producta, Bacteroides sp. OM08 11, Clostridiumclostridioforme, Collinsella intestinalis; The gut microbiota biomarker set for schizophrenia includes the following 13 gut microbial species: Lactococcus piscium, Bacillus cereus group, Bacteroides plebeius, Bifidobacterium pseudocatenulatum, Dialister sp. CAG 357, Megamonas funiformis, Anaerorotruncus colihominis, Bacteroides coprocola, Veillonella parvula, Akkermansia muciniphila, Prevotella sp. CAG 279, Mycobacterium avium, and Bifidobacterium longum; The intestinal microbial marker combination for colonic lesions phenotype included the following 35 intestinal microbial species: Parvimonasmicra, Mogibacterium diversum, Streptococcus mitis, Clostridium methylpentosum, Lawsonibacter asaccharolyticus, Butyricicoccus pullicaecorum, Veillonellaparvula, Allisonella histaminiformans, Collinsella aerofaciens, Enterococcusfaecalis, Acidaminococcus intestini, Prevotella sp. CAG 1092, Bifidobacteriumbreve, Fusobacterium ulcerans, Dialister pneumosintes, Alloscardovia omnicolens, Klebsiella aerogenes, Rothia mucilaginosa, Ruminococcus sp. CAG 403, Parabacteroides goldsteinii, Fusobacterium mortiferum, Streptococcus pasteurianus, Gemella morbillorum, Phascolarctobacterium sp. CAG 266, Prevotellaintermedia, Clostridium sp. CAG 242, Streptococcus sobrinus, Roseburiainulinivorans, Fusobacterium sp. oral taxon 370, Clostridium disporicum, Flavonifractor plautii, Odoribacter splanchnicus, Adlercreutzia equolifaciens, Eubacterium ramulus, Intestinimonas butyriciproducens; The gut microbiota biomarker combination for abnormal blood pressure phenotypes includes the following 18 gut microbial species: Actinomyces sp. ICM47, Proteobacteria bacterium CAG 139, Agathobaculum butyriciproducens, Clostridium sp. CAG 299, Oscillibacter sp. 57 20, Prevotella copri, Faecalibacterium prausnitzii, Lactobacillus sanfranciscensis, Roseburia intestinalis, Roseburia sp. CAG 471, Clostridium citroniae, Flavonifractorplautii, Actinomyces sp. oral taxon 181, Actinomyces odontolyticus, Klebsiella pneumoniae, Bilophila wadsworthia, Prevotella sp. AM42 24, Bacteroides eggerthii; The gut microbiota biomarker set for glucose intolerance includes the following 11 gut microbial species: Anaerostipes hadrus, Phascolarctobacterium faecium, Desulfovibrionaceaebacterium, Clostridium leptum, Prevotella stercorea, Alistipes inops, Parabacteroides distasonis, Intestinibacter bartlettii, Bacteroides nordii, and Lactobacillus acidophilus; The gut microbiota biomarker set for cirrhosis includes the following 20 gut microbial species: Lactobacillus salivarius, Prevotella stercorea, Prevotella buccae, Fusobacterium mortiferum, Veillonella atypica, Clostridium symbiosum, Veillonella sp. T110116, Aggregatibacter segnis, Lachnospira pectinoschiza, Dielma fastidiosa, Streptococcus vestibularis, Alistipes finegoldii, Streptococcus anginosusgroup, Tyzzerella nexilis, Lactobacillus fermentum, Eubacterium sp. CAG 251, Lactobacillus rogosae, Eubacterium eligens, Streptococcus infantarius, and Dorea sp. CAG 317. The melanoma phenotypic gut microbiota biomarker suite includes the following 20 gut microbial species: Actinomyces oris, Asaccharobacter celatus, Eggerthella lenta, Coprococcuseutactus, Bifidobacterium animalis, Faecalitalea cylindroides, Bifidobacterium dentium, Alistipes inops, Eubacterium sp. CAG 180, Bacteroides intestinalis, Methanobrevibacter smithii, Lactobacillus plantarum, Alistipes putredinis, Clostridium disporicum, Firmicutes bacterium CAG 94, Streptococcus anginosusgroup, Alistipes finegoldii, Eubacterium sp. CAG 251, Adlercreutziaequolifaciens, Actinomyces naeslundii; The gut microbiota markers for ankylosing spondylitis include the following 16 gut microbial species: Neisseria flavescens, Bifidobacterium pseudocatenulatum, Collinsella massiliensis, Pseudoflavonifractor sp. An184, Bacteroides plebeius, Bacteroides dorei, Rothia mucilaginosa, Prevotella copri, Roseburia faecis, Bacteroides intestinalis, Prevotella melaninogenica, Bacteroides stercoris, Clostridium citroniae, Prevotella stercorea, Barnesiella intestinihominis, and Bifidobacterium bifidum; The cardiovascular disease phenotype gut microbial marker panel includes the following 23 gut microbial species: Collinsella massiliensis, Eubacterium sp. OM08 24, Tyzzerella nexilis, Eubacterium sp. CAG 251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemaniafiliformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaera elsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacterium varium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonas virosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans; The gut microbiota markers for fatty liver phenotype include the following 18 gut microbial species: *Eubacterium rectale*, *Dorea* sp. CAG 317, *Anaerotignum lactatifermentans*, *Coprococcus comes*, *Eubacterium* sp. CAG 180, *Clostridium leptum*, *Acidaminococcus intestini*, *Streptococcus thermophilus*, *Dialister* sp. CAG 357, *Actinomyces* sp. oral taxon 181, *Eubacterium hallii*, *Anaeromassilibacillus* sp. An250, *Streptococcus parasanguinis*, *Streptococcus infantis*, *Eubacterium* sp. CAG 251, *Ruminococcus bicirculans*, *Bacteroides stercoris*, and *Methanobrevibacter smithii*.

[0013] The third technical problem to be solved by the present invention is to provide a chronic disease early screening kit, which includes reagents for detecting the abundance of at least one group of gut microbial markers among the above 12 groups of gut microbial marker combinations.

[0014] The fourth technical problem to be solved by this invention is to provide a method for constructing an early screening model for chronic diseases, the method comprising the following steps: 1) Using the above-mentioned screening methods for gut microbiota biomarkers of healthy individuals and chronic diseases, combinations of gut microbiota biomarkers for healthy phenotypes and various chronic disease phenotypes were obtained; 2) In the training set, for each phenotype, the combination of microbial biomarkers for that phenotype selected in step 1) is used as the model input features. The sample of that phenotype is used as the target sample, and the other samples are used as control samples. The random forest binary classification risk model is trained to generate the predicted probability of each sample belonging to that phenotype, that is, the risk index of that sample under that phenotype. 3) Using the risk indicators of all samples in the training set under each phenotype as input features, train a random forest multi-classification model to complete the construction of an early screening model for chronic diseases.

[0015] In step 2 above, it is preferable to balance the number of target samples and control samples using the SMOTE oversampling algorithm.

[0016] After completing the model construction in step 3 above, it is preferable to use the five-fold cross-validation method to evaluate the stability and discrimination ability of the early screening model for chronic diseases.

[0017] The fifth technical problem to be solved by the present invention is to provide an early screening model for chronic diseases constructed using the above method.

[0018] The sixth technical problem to be solved by the present invention is to provide the use of the above-mentioned early screening kit and model for chronic diseases, which can be used for early screening and risk assessment of one or more chronic diseases, including type 2 diabetes, inflammatory bowel disease, schizophrenia, colonic diseases, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver.

[0019] This invention achieves early screening and risk assessment for 11 common chronic diseases (including type 2 diabetes, inflammatory bowel disease, schizophrenia, colitis, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver) by screening gut microbiota biomarkers and establishing a multi-disease early screening model. Compared with existing screening technologies, this invention only requires detecting the abundance of specific microbial biomarkers in the subject's fecal sample to achieve early screening and risk assessment for these 11 common chronic diseases. It does not rely on imaging, is non-invasive, and has high coverage, accuracy, and specificity. Furthermore, it is convenient, efficient, and low-cost, making it particularly suitable for applications such as routine screening in health check-up centers, high-risk group early warning in community clinics, clinical follow-up monitoring of chronic disease patients, and personalized health management. It can achieve risk stratification and precise intervention for healthy individuals and sub-healthy populations. Attached Figure Description

[0020] Figure 1 It is the cross-validation result of the multi-disease early screening model for identifying healthy individuals.

[0021] Figure 2 This is the cross-validation result of the multi-disease early screening model for identifying patients with type 2 diabetes.

[0022] Figure 3 This is the cross-validation result of the multi-disease early screening model for identifying patients with inflammatory bowel disease.

[0023] Figure 4 This is the cross-validation result of the multi-disease early screening model for identifying patients with schizophrenia.

[0024] Figure 5 This is the cross-validation result of the multi-disease early screening model for identifying patients with colonic lesions.

[0025] Figure 6 This is the cross-validation result of the multi-disease early screening model for identifying patients with abnormal blood pressure.

[0026] Figure 7 This is the cross-validation result of the multi-disease early screening model for identifying patients with impaired glucose tolerance.

[0027] Figure 8 This is the cross-validation result of the multi-disease early screening model for identifying patients with cirrhosis.

[0028] Figure 9 This is the cross-validation result of the multi-disease early screening model for identifying melanoma patients.

[0029] Figure 10 This is the cross-validation result of the multi-disease early screening model for identifying patients with ankylosing spondylitis.

[0030] Figure 11 This is the cross-validation result of the multi-disease early screening model for identifying patients with cardiovascular diseases.

[0031] Figure 12 This is the cross-validation result of the multi-disease early screening model for identifying patients with fatty liver.

[0032] Figure 13 This is the test result of a multi-disease early screening model for identifying healthy individuals.

[0033] Figure 14 This is the test result of a multi-disease early screening model for identifying patients with type 2 diabetes.

[0034] Figure 15 This is the test result of a multi-disease early screening model for identifying patients with inflammatory bowel disease.

[0035] Figure 16 This is the test result of a multi-disease early screening model for identifying patients with schizophrenia.

[0036] Figure 17 This is the test result of a multi-disease early screening model for identifying patients with colonic lesions.

[0037] Figure 18 This is the test result of a multi-disease early screening model for identifying patients with abnormal blood pressure.

[0038] Figure 19 This is the test result of a multi-disease early screening model for identifying patients with impaired glucose tolerance.

[0039] Figure 20 This is the test result of a multi-disease early screening model for identifying patients with cirrhosis.

[0040] Figure 21 This is the test result of a multi-disease early screening model for identifying melanoma patients.

[0041] Figure 22This is the test result of a multi-disease early screening model for identifying patients with ankylosing spondylitis.

[0042] Figure 23 This is the test result of the multi-disease early screening model for identifying patients with cardiovascular diseases.

[0043] Figure 24 This is the test result of a multi-disease early screening model for identifying patients with fatty liver.

[0044] Figure 25 It is the external test result of the multi-disease early screening model for identifying healthy samples.

[0045] Figure 26 It is the external test result of the multi-disease early screening model for identifying patients with inflammatory bowel disease.

[0046] Figure 27 It is the external test result of the multi-disease early screening model for identifying patients with colonic lesions.

[0047] Figure 28 It is the external test result of the multi-disease early screening model for identifying patients with ankylosing spondylitis.

[0048] Figure 29 It is the external test result of the multi-disease early screening model for identifying patients with type 2 diabetes.

[0049] Figure 30 This is a comparison of the abundance of gut microbiota in clinical samples with that in healthy individuals.

[0050] Figure 31 This is the Shannon diversity index result of the gut microbiota in clinical samples. Detailed Implementation

[0051] To provide a more detailed understanding of the technical content, features, and effects of the present invention, the technical solution of the present invention will now be described in further detail with reference to the accompanying drawings.

[0052] Example 1: Construction of an early screening model for chronic diseases I. Data Collection Curated Metagenomic Data is a public metagenomic database built on high-quality metagenomic sequencing studies published internationally. This database adopts standardized data processing procedures, strict quality control strategies, and consistent species annotation systems for raw sequencing data. It features reliable data sources, standardized processing procedures, and high cross-cohort comparability, and is widely used in gut microbiota-related research.

[0053] In this embodiment, relative abundance data of gut microbiota metagenomics from all 93 cohorts in the CuratedMetagenomicData database, totaling 22,588 samples, were downloaded, along with corresponding clinical information (including sequencing platform, phenotype, sampling source, antibiotic use, body mass index (BMI), and age). The samples were then screened according to the following criteria: Illumina high-throughput sequencing platform was used; each cohort had at least 50 samples; samples with infectious or acute disease phenotypes were excluded; and only adult subjects with fecal samples, no antibiotic use records, and a BMI between 18.5 and 25 were retained. Ultimately, 7,801 samples from 49 cohorts were selected, covering healthy individuals and 11 common chronic disease phenotypes (including type 2 diabetes, inflammatory bowel disease, schizophrenia, colitis, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver).

[0054] II. Preprocessing of Metagenomic Relative Abundance Data The 7801 samples were randomly divided into a training set and a test set. The training set, containing 85% of the samples, was used for screening phenotypic microbial biomarkers, constructing risk indicators, and optimizing early screening model parameters. The test set, containing 15% of the samples, was used for independent evaluation of the constructed early screening model.

[0055] The relative abundance data of gut microbial metagenomics for each sample were normalized using a centered log-ratio (CLR) to eliminate the influence of sparsity and scaling effects on subsequent analyses.

[0056] III. Screening of Microbial Markers The training set sample data was randomly divided into a biomarker discovery set and a biomarker evaluation set at a ratio of 3:1.

[0057] In the biomarker discovery set, for each phenotype (including healthy status and 11 chronic disease phenotypes such as type 2 diabetes, inflammatory bowel disease, schizophrenia, colitis, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver), samples belonging to the target phenotype were used as target samples, and the remaining samples were used as control samples. A random subset of 60% of the control samples was randomly selected and combined with all target samples to form a random biomarker discovery subset for microbial biomarker screening analysis. For each random biomarker discovery subset, the SMOTE oversampling algorithm was used to maintain a balance between the number of target and control samples in each analysis, thereby improving the model's ability to identify minority categories. For each random biomarker discovery subset, a nested cross-validation strategy was adopted. The inner layer used three-fold cross-validation to determine the optimal regularization parameter λ of the LASSO regression model, and the outer layer constructed and trained the LASSO regression model under the optimal λ parameter condition. The regression coefficients corresponding to each microorganism in the LASSO regression model are extracted and used as phenotypic specificity scores of each microorganism's ability to discriminate the target phenotype, thus obtaining candidate microbial features and their phenotypic specificity ranking. The above process of constructing a subset of random biomarkers and scoring phenotypic specificity is repeated 100 times, and the average phenotypic specificity score of each microorganism in all repetitions is calculated as the basis for the final phenotypic specificity ranking of candidate microbial features.

[0058] Based on the phenotypic specificity ranking of candidate microbial features obtained above, a sequence-forward feature selection strategy was adopted to progressively stack microbial features according to their phenotypic specificity from high to low, constructing candidate microbial combinations of different sizes. For each candidate microbial combination, a random forest model was trained based on the balanced data, and the model was used to predict the disease probability of each sample. When the predicted probability of a candidate microbial combination for the target phenotypic sample was significantly higher than that for non-target phenotypic samples (U test p < 0.05), the candidate microbial combination was deemed to have phenotypic specificity discrimination ability and was defined as a candidate microbial biomarker combination with phenotypic specificity.

[0059] Each candidate microbial biomarker combination was independently evaluated using the biomarker evaluation set data. The method was as follows: Based on the candidate microbial biomarker combinations, a random forest method was used to distinguish between samples of the target phenotype and those of non-target phenotypes, and the predicted probability of each sample belonging to the target phenotype was output. An F1 score was calculated by combining the predicted probability with the true phenotype label, and the candidate microbial biomarker combination with the highest F1 score was selected as the final gut microbial biomarker combination for that phenotype. This process was performed independently for all phenotypes, ultimately obtaining the gut microbial biomarker combinations corresponding to each phenotype.

[0060] The gut microbiota markers for health status and 11 chronic disease phenotypes obtained from the final screening are as follows: The microbial marker profile of health status included the following 51 gut microbial species: Faecalibacteriumprausnitzii, Streptococcus thermophilus, Bifidobacterium angulatum, Finegoldiamagna, Pseudoflavonifractor capillosus, Coprobacter sp., Denitobacteriumdetoxificans, Prevotella sp. CAG 1124, Asaccharobacter celatus, Peptoniphiluslacrimalis, Anaerobiospirillum thomasii, Collinsella massiliensis, Prevotellatimonensis, Enorma massiliensis, Streptococcus infantarius, Butyricimonassynergistica, Ruminococcus sp. CAG 563, Bacteroides sp. D2, Massiliomicrobiotatimonensis, Bacteroides uniformis, Prevotella buccalis, Bacteroidespectinophilus, Bifidobacterium catenulatum, Bilophila wadsworthia, Sharpeaazabuensis, Porphyromonas sp. HMSC065F10, Bacteroides stercorirosoris, Cutibacterium acnes, Mitsuokella jalaludinii, Lachnoclostridium sp. An118, Leuconostoc gelidum, Blautia sp. CAG 257, Olsenella scatoligenes, Intestinimonas butyriciproductuns, Anaerotruncus sp. CAG 528, Citrobacter freundii, Actinomyces sp.oral taxon 448, Aeriscardovia aeriphila, Comamonas kerstersii, Alistipesshahii, Prevotella sp. CAG 520, Romboutsia ilealis, Clostridium clostridioforme, Megasphaera elsdenii, Escherichia marmotae, Victivallis vadensis, Catenibacterium mitsuokai, Lactococcus petauri, Clostridium sp. CAG 510, Megamonas hypermegale, Lactococcus garvieae.

[0061] The microbial marker set for type 2 diabetes includes the following 25 gut microbial species: Collinsellamassiliensis, Aeriscardovia aeriphila, Gardnerella vaginalis, Eubacterium sp.CAG 251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemania filiformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaera elsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacteriumvarium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonas virosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans.

[0062] The microbial marker composition of inflammatory bowel disease includes the following 27 intestinal microbial species: Enterorhabduscaecimuris, Enorma massiliensis, Bacteroides sp. CAG 144, Eisenbergiella massiliensis, Streptococcus lutetiensis, Anaeromassilibacillus sp. An250, Bacteroides vulgatus, Dorea longicatena, Lactobacillus sakei, Peptostreptococcus stomatis, Bacteroides sp. 43 108, Prevotella sp. 885, Fusobacterium nucleatum, Enterococcus faecium, Eubacterium sp. CAG 786, Blautia hansenii, Faecalibacterium prausnitzii, Blautia hydrogenotrophica, Lawsonibacterasaccharolyticus, Dialister pneumosintes, Oscillibacter sp. 57 20, Streptococcus mutans, Blautia obeum, Blautia producta, Bacteroides sp. OM08 11, Clostridium clostridioforme, Collinsella intestinalis.

[0063] The microbial biomarker combination for schizophrenia includes the following 13 gut microbial species: Lactococcus piscium, Bacillus cereus group, Bacteroides plebeius, Bifidobacteriumpseudocatenulatum, Dialister sp. CAG 357, Megamonas funiformis, Anaerorotruncus colihominis, Bacteroides coprocola, Veillonella parvula, Akkermansia muciniphila, Prevotella sp. CAG 279, Mycobacterium avium, and Bifidobacterium longum.

[0064] The microbial marker combination for colonic lesions included the following 35 intestinal microbial species: Parvimonasmicra, Mogibacterium diversum, Streptococcus mitis, Clostridium methylpentosum, Lawsonibacter asaccharolyticus, Butyricicoccus pullicaecorum, Veillonellaparvula, Allisonella histaminiformans, Collinsella aerofaciens, Enterococcusfaecalis, Acidaminococcus intestini, Prevotella sp. CAG 1092, Bifidobacteriumbreve, Fusobacterium ulcerans, Dialister pneumosintes, Alloscardovia omnicolens, Klebsiella aerogenes, Rothia mucilaginosa, Ruminococcus sp. CAG 403, Parabacteroides goldsteinii, Fusobacterium mortiferum, Streptococcus pasteurianus, Gemella morbillorum, Phascolarctobacterium sp. CAG 266, Prevotellaintermedia, Clostridium sp. CAG 242, Streptococcus sobrinus, Roseburiainulinivorans, Fusobacterium sp. oral taxon 370, Clostridium disporicum, Flavonifractor plautii, Odoribacter splanchnicus, Adlercreutzia equolifaciens, Eubacterium ramulus, Intestinimonas butyriciproducens.

[0065] The microbial biomarker combination for abnormal blood pressure includes the following 18 gut microbiota species: Actinomyces sp. ICM47, Proteobacteria bacterium CAG 139, Agathobaculum butyriciproducens, Clostridium sp. CAG 299, Oscillibacter sp. 57 20, Prevotella copri, Faecalibacterium prausnitzii, Lactobacillus sanfranciscensis, Roseburia intestinalis, Roseburia sp. CAG 471, Clostridium citroniae, Flavonifractorplautii, Actinomyces sp. oral taxon 181, Actinomyces odontolyticus, Klebsiella pneumoniae, Bilophila wadsworthia, Prevotella sp. AM42 24, and Bacteroides eggerthii.

[0066] The microbial biomarker combination for impaired glucose tolerance includes the following 11 gut microbial species: Anaerostipeshadrus, Phascolarctobacterium faecium, Desulfovibrionaceae bacterium, Clostridium leptum, Prevotella stercorea, Alistipes inops, Parabacteroides distasonis, Intestinibacter bartlettii, Bacteroides nordii, and Lactobacillus acidophilus.

[0067] The microbial marker set for cirrhosis included the following 20 gut microbial species: Lactobacillus salivarius, Prevotella stercorea, Prevotella buccae, Fusobacterium mortiferum, Veillonella atypica, Clostridium symbiosum, Veillonella sp. T11011 6, Aggregatibacter segnis, Lachnospira pectinoschiza, Dielma fastidiosa, Streptococcus vestibularis, Alistipes finegoldii, Streptococcus anginosusgroup, Tyzzerella nexilis, Lactobacillus fermentum, Eubacterium sp. CAG 251, Lactobacillus rogosae, Eubacterium eligens, Streptococcus infantarius, Dorea sp. CAG 317.

[0068] Black and white microorganism markers include the following 20 kinds of microorganisms: Actinomycesoris, Asaccharobacter celatus, Eggerthella lenta, Coprococcus eutactus, Bifidobacterium animalis, Faecalitalea cylindroides, Bifidobacterium dentium, Alistipes inops, Eubacterium sp. CAG 180, Bacteroides intestinalis, Methanobrevibacter smithii, Lactobacillus plantarum, Alistipes putredinis, Clostridium disporicum, Firmicutes bacterium CAG 94, Streptococcus anginosusgroup, Alistipes finegoldii, Eubacterium sp. CAG 251,Adlercreutziaequolifaciens,Actinomyces naeslundii。

[0069] The microbial biomarker combination for ankylosing spondylitis includes the following 16 gut microbial species: Neisseria flamescens, Bifidobacterium pseudocatenulatum, Collinsella massiliensis, Pseudoflavonifractor sp. An184, Bacteroides plebeius, Bacteroides dorei, Rothia mucilaginosa, Prevotella copri, Roseburia faecis, Bacteroides intestinalis, Prevotella melaninogenica, Bacteroides stercoris, Clostridium citroniae, Prevotella stercorea, Barnesiella intestinihominis, and Bifidobacterium bifidum.

[0070] The microbial marker portfolio for cardiovascular disease included the following 23 gut microbial species: Collinsellamassiliensis, Eubacterium sp. OM08 24, Tyzzerella nexilis, Eubacterium sp. CAG251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemania filiformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaera elsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacteriumvarium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonas virosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans.

[0071] The microbial biomarker combination for fatty liver includes the following 18 gut microbial species: *Eubacterium rectale*, *Dorea* sp. CAG 317, *Anaerotignum lactatifermentans*, *Coprococcus comes*, *Eubacterium* sp. CAG 180, *Clostridium leptum*, *Acidaminococcus intestini*, *Streptococcus thermophilus*, *Dialister* sp. CAG 357, *Actinomyces* sp. oral taxon 181, *Eubacterium hallii*, *Anaeromassilibacillus* sp. An250, *Streptococcus parasanguinis*, *Streptococcus infantis*, *Eubacterium* sp. CAG 251, *Ruminococcus bicirculans*, *Bacteroides stercoris*, and *Methanobrevibacter smithii*.

[0072] IV. Construction and Evaluation of Multi-Disease Early Screening Models Based on the specific microbial biomarker combinations selected for each phenotype, early screening risk indicators for health status and 11 chronic diseases were constructed, and a multi-disease early screening model was built upon these indicators. The specific methods are as follows: In the training set, for each phenotype, a specific combination of microbial biomarkers corresponding to that phenotype is used as input features. Samples belonging to that phenotype are defined as target samples, and the remaining samples are defined as control samples. The SMOTE oversampling algorithm is used to maintain a balance between target and control samples across folds, improving the reliability of the risk indicator under class imbalance conditions. Then, random forest binary classification risk modeling is performed based on the input features (i.e., a random forest binary classification risk model is constructed for each phenotype), and five-fold cross-validation is used to repeatedly evaluate the modeling process to measure the stability and discriminative power of the risk indicator. The predicted probability of each sample belonging to the target phenotype is output, i.e., the risk indicator of that sample under the corresponding phenotype. The risk indicator can quantify an individual's disease risk level under a corresponding phenotype.

[0073] After obtaining the risk indicators corresponding to each phenotype, the risk indicators of all samples in the training set under each phenotype are used as input features. The random forest method is used for multi-class classification modeling to complete the construction of a multi-disease early screening model, which can simultaneously predict the risk distribution of samples in healthy state and 11 chronic disease states.

[0074] To obtain robust performance evaluation results, this embodiment employs five-fold cross-validation to evaluate the modeling process of the aforementioned multi-disease early screening model. The specific method is as follows: In each fold, performance testing is performed using validation samples that were not used in training at that fold. By comparing the output results of the validation samples with the true phenotypic labels, the area under the receiver operating characteristic (AUC) curve corresponding to each phenotype is calculated to quantify the ability of the multi-disease early screening model to distinguish between target and non-target samples under that phenotype. The cross-validation AUC results for each phenotype are shown below. Figures 1-12 As shown. Figure 1 For example, Figure 1 As shown, the AUC of 0.85 for the multi-disease early screening model in identifying healthy samples indicates that the method can distinguish between healthy and unhealthy samples with high accuracy, thus supporting early identification and risk stratification of health status. The AUC results for other phenotypes are similar. A higher AUC value indicates a stronger discriminative ability for that phenotype. The AUC results obtained from five-fold cross-validation show that the early screening risk index and the multi-disease early screening model constructed in this embodiment have generally stable discriminative ability and good reproducibility in multi-phenotype scenarios.

[0075] V. Performance Testing of Multi-Disease Early Screening Model 1. Performance testing of the multi-disease early screening model on the test set. The 12 gut microbiota biomarkers selected in step three, targeting health status and 11 common chronic diseases, were used to test and evaluate the multi-disease early screening model constructed in step four on a test set (without participating in any model training or parameter optimization). The specific method is as follows: Using a random forest binary classification risk model of the 12 phenotypes constructed in step four, the risk index corresponding to each sample in the test set under the 12 phenotypes was calculated sequentially. This risk index was then used as input features and fed into the multi-disease early screening model to obtain the risk prediction results corresponding to each phenotype. By comparing the output results of the multi-disease early screening model with the true phenotype labels of the samples in the test set, the area under the receiver operating characteristic (AUC) curve corresponding to each phenotype was calculated to evaluate the generalization performance of the gut-specific microbiota biomarker combination and the multi-disease early screening model in independent samples.

[0076] Test evaluation results as follows Figures 13-24 As shown, with Figure 13 For example, Figure 13The test results corresponding to the healthy status are shown, with an AUC value of 0.83, indicating that the multi-disease early screening model constructed in this embodiment can still effectively distinguish between healthy and unhealthy individuals in an independent population, demonstrating good generalization ability and practical application potential. The test results for other chronic disease phenotypes are similar, all showing stable AUCs on the test set, indicating that the gut microbiota biomarkers screened in this embodiment and the risk indicators constructed based on them have good cross-sample robustness and applicability for multi-disease early screening.

[0077] 2. Performance testing of the multi-disease early screening model on an external test dataset. To further verify the generalization ability and robustness of the gut microbiota biomarkers selected and the constructed multi-disease early screening model in different population data, this embodiment introduces an external test dataset independent of the training and test sets mentioned above for validation. This external dataset is derived from metagenomic relative abundance data directly downloaded from the Gut Microbiome Wellness Index2 (GMWI2) study. After strictly removing samples that are duplicates or have overlapping sources with those in the CuratedMetagenomicData database, 3606 samples were ultimately retained, covering 5 phenotypes, including 2596 healthy samples, 457 inflammatory bowel disease patients, 264 colon disease patients, 151 ankylosing spondylitis patients, and 138 type 2 diabetes patients.

[0078] The relative abundance data of metagenomics for these samples were also normalized using the central logarithmic transformation (CLR). Subsequently, based on the phenotype-specific gut microbiota biomarkers screened in this embodiment, the risk indicators for each sample in the external test set under the healthy state and the above four chronic disease phenotypes were calculated sequentially using the random forest binary classification risk models constructed in step four above for the five phenotypes: healthy, inflammatory bowel disease, colitis, ankylosing spondylitis, and type 2 diabetes. These risk indicators were then input into the multi-disease early screening model for prediction and evaluation. The results are as follows: Figures 25-29 As shown, the multi-disease early screening model maintained stable discrimination performance in external datasets, indicating that the gut microbiota biomarkers screened in this embodiment have good cross-cohort generalization ability and repeatability, and can be stably applied in real population data from different sources, providing reliable support for early screening and risk assessment of chronic diseases.

[0079] 3. Performance testing of multi-disease early screening models using clinical samples. To further test the applicability and individual-level interpretability of the multi-disease early screening model constructed in this embodiment in real clinical scenarios, a sample with rich clinical information was selected. This sample came from a 51-year-old male patient diagnosed with cirrhosis through tissue biopsy and imaging. Analysis of the patient's gut microbiota composition based on the relative abundance data of fecal metagenomics revealed that, compared to healthy controls, this sample exhibited a microbial imbalance significantly associated with cirrhosis, including… Figure 30 The relative abundance of species such as *Aggregatibacter segnis*, *Lactobacillus salivarius*, and *Streptococcus vestibularis* increased, while the relative abundance of health-related species such as *Alistipes finegoldii* and *Eubacterium sp. CAG 251* decreased significantly. This is consistent with the aforementioned changes in gut microbiota composition. Figure 31 The sample's gut microbial alpha diversity (Shannon index) was significantly lower than the healthy baseline level, indicating that its gut ecosystem was in a state of significant disturbance.

[0080] Building upon this, the random forest binary classification risk model for the 12 phenotypes constructed in step four was used to further calculate the risk indicators for this sample under healthy and 11 common chronic disease phenotypes, and an individualized risk spectrum was generated using a multi-disease early screening model. The results showed that the multi-disease early screening model predicted the highest risk phenotype for this sample as cirrhosis (57%), followed by healthy status (29%) and inflammatory bowel disease (4%). Based on this risk distribution, the multi-disease early screening model indicated a high risk of cirrhosis and recommended targeted liver-related examinations. Notably, the patient's routine clinical indicators did not show significant abnormalities; their serum albumin (38.9 g / L), creatinine (78 μmol / L), and body mass index (21.45 kg / m²) were normal. 2 All levels were within the normal reference range, with only a slight increase in total bilirubin. This indicates that even when routine clinical tests are insensitive or show no obvious abnormalities, the gut microbiota-based multi-disease early screening model in this embodiment can still capture potential disease risk signals. Clinical pathology and imaging examinations also confirmed that the patient did indeed have cirrhosis, further highlighting the limitations of traditional indicators in identifying subclinical stages and the potential advantages of microbiome methods in early disease identification.

[0081] The above embodiments are merely feasible or preferred embodiments of the present invention and are used to illustrate the present invention. They are not intended to limit the scope of the patent application of the present invention. Therefore, all equivalent changes and modifications made in accordance with the scope of the patent application of the present invention should fall within the scope of the patent application of the present invention.

Claims

1. A method for screening gut microbiota biomarkers for healthy individuals and those with chronic diseases, characterized in that, Includes the following steps: 1) Obtain relative abundance data of gut microbiota metagenomics and clinical information from the samples; 2) Screening samples that cover both healthy and chronic disease phenotypes; 3) Randomly divide the samples selected in step 2) into training set and test set, and normalize the relative abundance data of gut microbial metagenomics for each sample; 4) In the training set, nested cross-validation and sequence forward feature selection are used to screen phenotype-specific gut microbial biomarkers for the healthy phenotype and each type of chronic disease phenotype. Specific steps include: randomly dividing the training set sample data into a biomarker discovery set and a biomarker evaluation set; in the biomarker discovery set, one type of phenotype samples are used as target samples, and the remaining samples are used as control samples. A subset of samples is randomly selected from the control samples and combined with the target samples to form a random biomarker discovery subset for that phenotype; for each random biomarker discovery subset, nested cross-validation is used to screen candidate microbial biomarkers that can specifically distinguish the corresponding phenotype. The phenotypic specificity scores of each microorganism were analyzed, and the above process was repeated multiple times. The average phenotypic specificity score of each microorganism across all replicates was calculated and ranked. Then, a sequence forward feature selection method was used to construct candidate microbial combinations of different sizes. A random forest model was trained to predict the probability of each phenotype for each sample. When the overall predicted probability of a candidate microbial combination for the target phenotype sample was significantly higher than that for non-target phenotype samples, the candidate microbial combination was determined as the candidate microbial biomarker combination for the target phenotype. Each candidate microbial biomarker combination was independently evaluated using a biomarker evaluation set, and finally, the gut microbial biomarker combination corresponding to each phenotype was obtained.

2. The method according to claim 1, characterized in that, The chronic diseases include one or more of the following: type 2 diabetes, inflammatory bowel disease, schizophrenia, colitis, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver.

3. The method according to claim 1, characterized in that, Step 2), the screening criteria include: using Illumina high-throughput sequencing, a cohort sample size of no less than 50, samples from non-infectious and acute diseases, and samples from feces of adults with no record of antibiotic use and a body mass index between 18.5 and 25.

4. The method according to claim 1, characterized in that, Step 3), the ratio of the sample size of the training set to the test set is 8.5:1.5; the normalization process adopts the central logarithmic transformation method.

5. The method according to claim 1, characterized in that, Step 4), the nested cross-validation method, uses three-fold cross-validation in the inner layer to determine the optimal regularization parameter of the LASSO regression model, and in the outer layer, under the condition of the optimal regularization parameter, constructs and trains the LASSO regression model. The regression coefficients corresponding to each microorganism in the LASSO regression model are the phenotypic specificity scores of each microorganism's ability to distinguish the target phenotype.

6. The method according to claim 1, characterized in that, Step 4), the sequence forward feature selection method gradually superimposes microbial features from high to low phenotypic specificity to construct candidate microbial combinations of different scales.

7. The method according to claim 1, characterized in that, Step 4) The method of independently evaluating each candidate microbial biomarker combination using a biomarker evaluation set includes the following steps: Based on the candidate microbial biomarker combination, a random forest method is used to distinguish between samples of the target phenotype and those of non-target phenotypes, and the predicted probability of each sample belonging to the target phenotype is output; the F1 score is calculated by combining the predicted probability with the true phenotype label, and the candidate microbial biomarker combination with the highest F1 score is selected as the final gut microbial biomarker combination for that phenotype.

8. The method according to any one of claims 1-7, characterized in that, Step 4) The combinations of gut microbiota markers obtained from screening are as follows: The healthy phenotype gut microbial marker set includes the following 51 gut microbial species: Faecalibacteriumprausnitzii, Streptococcus thermophilus, Bifidobacterium angulatum, Finegoldiamagna, Pseudoflavonifractor capillosus, Coprobacter sp., Denitobacteriumdetoxificans, Prevotella sp. CAG 1124, Asaccharobacter celatus, Peptoniphiluslacrimalis, Anaerobiospirillum thomasii, Collinsella massiliensis, Prevotellatimonensis, Enorma massiliensis, Streptococcus infantarius, Butyricimonassynergistica, Ruminococcus sp. CAG 563, Bacteroides sp. D2, Massiliomicrobiotatimonensis, Bacteroides uniformis, Prevotella buccalis, Bacteroidespectinophilus, Bifidobacterium catenulatum, Bilophila wadsworthia, Sharpeaazabuensis, Porphyromonas sp. HMSC065F10, Bacteroides stercorirosoris, Cutibacterium acnes, Mitsuokella jalaludinii, Lachnoclostridium sp. An118, Leuconostoc gelidum, Blautia sp. CAG 257, Olsenella scatoligenes, Intestinimonas butyriciproductuns, Anaerotruncus sp. CAG 528, Citrobacter freundii, Actinomyces sp.oral taxon 448, Aeriscardovia aeriphila, Comamonas kerstersii, Alistipesshahii, Prevotella sp. CAG 520, Romboutsia ilealis, Clostridium clostridioforme, Megasphaera elsdenii, Escherichia marmotae, Victivallis vadensis, Catenibacterium mitsuokai, Lactococcus petauri, Clostridium sp. CAG 510, Megamonas hypermegale, Lactococcus garvieae;. The gut microbiota biomarker set for type 2 diabetes includes the following 25 gut microbiome species: Collinsella mastiliensis, Aeriscardovia aeriphila, Gardnerella vaginalis, Eubacterium sp. CAG 251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemania filiformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaeraelsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacterium varium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonas virosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans; The intestinal microbial biomarker set for inflammatory bowel disease phenotypes includes the following 27 intestinal microbial species: Enterorhabdus caecimuris, Enorma massiliensis, Bacteroides sp. CAG 144, Eisenbergiella massiliensis, Streptococcus lutetiensis, Anaeromassilibacillus sp. An250, Bacteroides vulgatus, Dorea longicatena, Lactobacillus sakei, Peptostreptococcus stomatis, Bacteroides sp. 43 108, Prevotella sp. 885, Fusobacterium nucleatum, Enterococcus faecium, Eubacterium sp. CAG 786, Blautia hansenii, Faecalibacterium prausnitzii, Blautia hydrogenotrophica, Lawsonibacterasaccharolyticus, and Dialister. pneumosintes, Oscillibacter sp. 57 20, Streptococcusmutans, Blautia obeum, Blautia producta, Bacteroides sp. OM08 11, Clostridiumclostridioforme, Collinsella intestinalis; The gut microbiota biomarker set for schizophrenia includes the following 13 gut microbial species: Lactococcus piscium, Bacillus cereus group, Bacteroides plebeius, Bifidobacteriumpseudocatenulatum, Dialister sp. CAG 357, Megamonas funiformis, Anaerorotruncus colihominis, Bacteroides coprocola, Veillonella parvula, Akkermansia muciniphila, Prevotella sp. CAG 279, Mycobacterium avium, and Bifidobacterium longum; The intestinal microbial marker combination for colonic lesions phenotype included the following 35 intestinal microbial species: Parvimonasmicra, Mogibacterium diversum, Streptococcus mitis, Clostridium methylpentosum, Lawsonibacter asaccharolyticus, Butyricicoccus pullicaecorum, Veillonellaparvula, Allisonella histaminiformans, Collinsella aerofaciens, Enterococcusfaecalis, Acidaminococcus intestini, Prevotella sp. CAG 1092, Bifidobacteriumbreve, Fusobacterium ulcerans, Dialister pneumosintes, Alloscardovia omnicolens, Klebsiella aerogenes, Rothia mucilaginosa, Ruminococcus sp. CAG 403, Parabacteroides goldsteinii, Fusobacterium mortiferum, Streptococcus pasteurianus, Gemella morbillorum, Phascolarctobacterium sp. CAG 266, Prevotellaintermedia, Clostridium sp. CAG 242, Streptococcus sobrinus, Roseburiainulinivorans, Fusobacterium sp. oral taxon 370, Clostridium disporicum, Flavonifractor plautii, Odoribacter splanchnicus, Adlercreutzia equolifaciens, Eubacterium ramulus, Intestinimonas butyriciproducens; The gut microbiota biomarker combination for abnormal blood pressure phenotypes includes the following 18 gut microbial species: Actinomyces p. ICM47, Proteobacteria bacterium CAG 139, Agathobacillus butyriciproducens, Clostridium sp. CAG 299, Oscillibacter sp. 57 20, Prevotella copri, Faecalibacterium prausnitzii, Lactobacillus sanfranciscensis, Roseburia intestinalis, Roseburia sp. CAG 471, Clostridium citroniae, Flavonifractorplautii, Actinomyces sp. oral taxon 181, Actinomyces odontolyticus, Klebsiella pneumoniae, Bilophila wadsworthia, Prevotella sp. AM42 24, Bacteroides eggerthii; The gut microbiota biomarker set for glucose intolerance includes the following 11 gut microbial species: Anaerostipeshadrus, Phascolarctobacterium faecium, Desulfovibrionaceae bacterium, Clostridium leptum, Prevotella stercorea, Alistipes inops, Parabacteroides distasonis, Intestinibacter bartlettii, Bacteroides nordii, and Lactobacillus acidophilus; The liver cirrhosis phenotype gut microbial marker panel includes the following 20 gut microbial species: Lactobacillus salivarius, Prevotella stercorea, Prevotella buccae, Fusobacterium mortiferum, Veillonella atypica, Clostridium symbiosum, Veillonella sp. T11011 6, Aggregatibacter segnis, Lachnospira pectinoschiza, Dielma fastidiosa, Streptococcus vestibularis, Alistipes finegoldii, Streptococcus anginosusgroup, Tyzzerella nexilis, Lactobacillus fermentum, Eubacterium sp. CAG 251, Lactobacillus rogosae, Eubacterium eligens, Streptococcus infantarius, Dorea sp.CAG 317; The melanoma phenotypic gut microbiota biomarker set includes the following 20 gut microbial species: Actinomycesoris, Asaccharobacter celatus, Eggerthella lenta, Coprococcus eutactus, Bifidobacterium animalis, Faecalitalea cylindroides, Bifidobacterium dentium, Alistipes inops, Eubacterium sp. CAG 180, Bacteroides intestinalis, Methanobrevibacter smithii, Lactobacillus plantarum, Alistipes putredinis, Clostridium disporicum, Firmicutes bacterium CAG 94, Streptococcus anginosusgroup, Alistipes finegoldii, Eubacterium sp. CAG 251, Adlercreutziaequolifaciens, Actinomyces naeslundii; The ankylosing spondylitis phenotype gut microbiota biomarker set includes the following 16 gut microbial species: Neisseria flamescens, Bifidobacterium pseudocatenulatum, Collinsella massiliensis, Pseudoflavonifractor sp. An184, Bacteroides plebeius, Bacteroides dorei, Rothia mucilaginosa, Prevotella copri, Roseburia faecis, Bacteroides intestinalis, Prevotella melaninogenica, Bacteroides stercoris, Clostridium citroniae, Prevotella stercorea, Barnesiella intestinihominis, and Bifidobacterium bifidum; The gut microbiota biomarker combination for cardiovascular disease phenotypes includes the following 23 gut microbial species: Collinsella massiliensis, Eubacterium sp. OM08 24, Tyzzerella nexilis, Eubacterium sp. CAG251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemania filiformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaeraelsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacterium varium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonasvirosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans; The gut microbiota markers for fatty liver phenotype include the following 18 gut microbial species: *Eubacterium rectale*, *Dorea* sp. CAG 317, *Anaerotignum lactatifermentans*, *Coprococcus comes*, *Eubacterium* sp. CAG 180, *Clostridium leptum*, *Acidaminococcus intestini*, *Streptococcus thermophilus*, *Dialister* sp. CAG 357, *Actinomyces* sp. oral taxon 181, *Eubacterium hallii*, *Anaeromassilibacillus* sp. An250, *Streptococcus parasanguinis*, *Streptococcus infantis*, *Eubacterium* sp. CAG 251, *Ruminococcus bicirculans*, *Bacteroides stercoris*, and *Methanobrevibacter smithii*.

9. A combination of gut microbiota markers for healthy and chronic diseases obtained by screening using the method described in any one of claims 1-7, characterized in that: The healthy phenotype gut microbial marker set includes the following 51 gut microbial species: Faecalibacteriumprausnitzii, Streptococcus thermophilus, Bifidobacterium angulatum, Finegoldiamagna, Pseudoflavonifractor capillosus, Coprobacter sp., Denitobacteriumdetoxificans, Prevotella sp. CAG 1124, Asaccharobacter celatus, Peptoniphiluslacrimalis, Anaerobiospirillum thomasii, Collinsella massiliensis, Prevotellatimonensis, Enorma massiliensis, Streptococcus infantarius, Butyricimonassynergistica, Ruminococcus sp. CAG 563, Bacteroides sp. D2, Massiliomicrobiotatimonensis, Bacteroides uniformis, Prevotella buccalis, Bacteroidespectinophilus, Bifidobacterium catenulatum, Bilophila wadsworthia, Sharpeaazabuensis, Porphyromonas sp. HMSC065F10, Bacteroides stercorirosoris, Cutibacterium acnes, Mitsuokella jalaludinii, Lachnoclostridium sp. An118, Leuconostoc gelidum, Blautia sp. CAG 257, Olsenella scatoligenes, Intestinimonas butyriciproductuns, Anaerotruncus sp. CAG 528, Citrobacter freundii, Actinomyces sp.oral taxon 448, Aeriscardovia aeriphila, Comamonas kerstersii, Alistipesshahii, Prevotella sp. CAG 520, Romboutsia ilealis, Clostridium clostridioforme, Megasphaera elsdenii, Escherichia marmotae, Victivallis vadensis, Catenibacterium mitsuokai, Lactococcus petauri, Clostridium sp. CAG 510, Megamonas hypermegale, Lactococcus garvieae;. The gut microbial marker combination for type 2 diabetes phenotype includes the following 25 gut microbial species: Collinsellamassiliensis, Aeriscardovia aeriphila, Gardnerella vaginalis, Eubacterium sp.CAG 251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemania filiformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaera elsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacteriumvarium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonas virosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans; The intestinal microbial biomarker set for inflammatory bowel disease phenotypes includes the following 27 intestinal microbial species: Enterorhabdus caecimuris, Enorma massiliensis, Bacteroides sp. CAG 144, Eisenbergiella massiliensis, Streptococcus lutetiensis, Anaeromassilibacillus sp. An250, Bacteroides vulgatus, Dorea longicatena, Lactobacillus sakei, Peptostreptococcus stomatis, Bacteroides sp. 43 108, Prevotella sp. 885, Fusobacterium nucleatum, Enterococcus faecium, Eubacterium sp. CAG 786, Blautia hansenii, Faecalibacterium prausnitzii, Blautia hydrogenotrophica, Lawsonibacterasaccharolyticus, and Dialister. pneumosintes, Oscillibacter sp. 57 20, Streptococcusmutans, Blautia obeum, Blautia producta, Bacteroides sp. OM08 11, Clostridiumclostridioforme, Collinsella intestinalis; The gut microbiota biomarker set for schizophrenia includes the following 13 gut microbial species: Lactococcus piscium, Bacillus cereus group, Bacteroides plebeius, Bifidobacteriumpseudocatenulatum, Dialister sp. CAG 357, Megamonas funiformis, Anaerorotruncus colihominis, Bacteroides coprocola, Veillonella parvula, Akkermansia muciniphila, Prevotella sp. CAG 279, Mycobacterium avium, and Bifidobacterium longum; The intestinal microbial marker combination for colonic lesions phenotype included the following 35 intestinal microbial species: Parvimonasmicra, Mogibacterium diversum, Streptococcus mitis, Clostridium methylpentosum, Lawsonibacter asaccharolyticus, Butyricicoccus pullicaecorum, Veillonellaparvula, Allisonella histaminiformans, Collinsella aerofaciens, Enterococcusfaecalis, Acidaminococcus intestini, Prevotella sp. CAG 1092, Bifidobacteriumbreve, Fusobacterium ulcerans, Dialister pneumosintes, Alloscardovia omnicolens, Klebsiella aerogenes, Rothia mucilaginosa, Ruminococcus sp. CAG 403, Parabacteroides goldsteinii, Fusobacterium mortiferum, Streptococcus pasteurianus, Gemella morbillorum, Phascolarctobacterium sp. CAG 266, Prevotellaintermedia, Clostridium sp. CAG 242, Streptococcus sobrinus, Roseburiainulinivorans, Fusobacterium sp. oral taxon 370, Clostridium disporicum, Flavonifractor plautii, Odoribacter splanchnicus, Adlercreutzia equolifaciens, Eubacterium ramulus, Intestinimonas butyriciproducens; The gut microbiota biomarker combination for abnormal blood pressure phenotypes includes the following 18 gut microbial species: Actinomyces p. ICM47, Proteobacteria bacterium CAG 139, Agathobacillus butyriciproducens, Clostridium sp. CAG 299, Oscillibacter sp. 57 20, Prevotella copri, Faecalibacterium prausnitzii, Lactobacillus sanfranciscensis, Roseburia intestinalis, Roseburia sp. CAG 471, Clostridium citroniae, Flavonifractorplautii, Actinomyces sp. oral taxon 181, Actinomyces odontolyticus, Klebsiella pneumoniae, Bilophila wadsworthia, Prevotella sp. AM42 24, Bacteroides eggerthii; The gut microbiota biomarker set for glucose intolerance includes the following 11 gut microbial species: Anaerostipeshadrus, Phascolarctobacterium faecium, Desulfovibrionaceae bacterium, Clostridium leptum, Prevotella stercorea, Alistipes inops, Parabacteroides distasonis, Intestinibacter bartlettii, Bacteroides nordii, and Lactobacillus acidophilus; The liver cirrhosis phenotype gut microbial marker panel includes the following 20 gut microbial species: Lactobacillus salivarius, Prevotella stercorea, Prevotella buccae, Fusobacterium mortiferum, Veillonella atypica, Clostridium symbiosum, Veillonella sp. T11011 6, Aggregatibacter segnis, Lachnospira pectinoschiza, Dielma fastidiosa, Streptococcus vestibularis, Alistipes finegoldii, Streptococcus anginosusgroup, Tyzzerella nexilis, Lactobacillus fermentum, Eubacterium sp. CAG 251, Lactobacillus rogosae, Eubacterium eligens, Streptococcus infantarius, Dorea sp.CAG 317; The melanoma phenotypic gut microbiota biomarker set includes the following 20 gut microbial species: Actinomycesoris, Asaccharobacter celatus, Eggerthella lenta, Coprococcus eutactus, Bifidobacterium animalis, Faecalitalea cylindroides, Bifidobacterium dentium, Alistipes inops, Eubacterium sp. CAG 180, Bacteroides intestinalis, Methanobrevibacter smithii, Lactobacillus plantarum, Alistipes putredinis, Clostridium disporicum, Firmicutes bacterium CAG 94, Streptococcus anginosusgroup, Alistipes finegoldii, Eubacterium sp. CAG 251, Adlercreutziaequolifaciens, Actinomyces naeslundii; The ankylosing spondylitis phenotype gut microbiota biomarker set includes the following 16 gut microbial species: Neisseria flamescens, Bifidobacterium pseudocatenulatum, Collinsella massiliensis, Pseudoflavonifractor sp. An184, Bacteroides plebeius, Bacteroides dorei, Rothia mucilaginosa, Prevotella copri, Roseburia faecis, Bacteroides intestinalis, Prevotella melaninogenica, Bacteroides stercoris, Clostridium citroniae, Prevotella stercorea, Barnesiella intestinihominis, and Bifidobacterium bifidum; The gut microbiota biomarker combination for cardiovascular disease phenotypes includes the following 23 gut microbial species: Collinsella massiliensis, Eubacterium sp. OM08 24, Tyzzerella nexilis, Eubacterium sp. CAG251, Lactobacillus salivarius, Streptococcus oralis, Klebsiella pneumoniae, Bacteroides plebeius, Eubacterium ventriosum, Holdemania filiformis, Streptococcus mitis, Clostridium bolteae, Gordonibacter pamelaeae, Megasphaeraelsdenii, Streptococcus vestibularis, Bacteroides intestinalis, Fusobacterium varium, Clostridium bolteae CAG 59, Clostridium sp. CAG 242, Butyricimonasvirosa, Phascolarctobacterium faecium, Ruminococcus callidus, Ruthenibacterium lactatiformans; The gut microbiota markers for fatty liver phenotype include the following 18 gut microbial species: *Eubacterium rectale*, *Dorea* sp. CAG 317, *Anaerotignum lactatifermentans*, *Coprococcus comes*, *Eubacterium* sp. CAG 180, *Clostridium leptum*, *Acidaminococcus intestini*, *Streptococcus thermophilus*, *Dialister* sp. CAG 357, *Actinomyces* sp. oral taxon 181, *Eubacterium hallii*, *Anaeromassilibacillus* sp. An250, *Streptococcus parasanguinis*, *Streptococcus infantis*, *Eubacterium* sp. CAG 251, *Ruminococcus bicirculans*, *Bacteroides stercoris*, and *Methanobrevibacter smithii*.

10. A chronic disease early screening kit, characterized in that, It includes reagents for detecting the abundance of gut microbial markers for at least one phenotype of claim 9.

11. A method for constructing an early screening model for chronic diseases, characterized in that, Includes the following steps: 1) Using the method described in any one of claims 1-8, a combination of gut microbiota markers for healthy phenotypes and various chronic disease phenotypes is obtained by screening; 2) In the training set, for each phenotype, the combination of microbial biomarkers for that phenotype selected in step 1) is used as the model input features. The sample of that phenotype is used as the target sample, and the other samples are used as control samples. The random forest binary classification risk model is trained to generate the predicted probability of each sample belonging to that phenotype, that is, the risk index of that sample under that phenotype. 3) Using the risk indicators of all samples in the training set under each phenotype as input features, train a random forest multi-classification model to complete the construction of an early screening model for chronic diseases.

12. The method according to claim 11, characterized in that, Step 2) Balance the number of target samples and control samples using the SMOTE oversampling algorithm; Step 3) After completing the model construction, use the five-fold cross-validation method to evaluate the stability and discrimination ability of the early screening model for chronic diseases.

13. The application of the kit of claim 10 in early screening for chronic diseases, wherein the chronic diseases include one or more of type 2 diabetes, inflammatory bowel disease, schizophrenia, colitis, abnormal blood pressure, impaired glucose tolerance, cirrhosis, melanoma, ankylosing spondylitis, cardiovascular disease, and fatty liver.

Images