Diabetes Prediction System and Model Construction Method Based on Gut Microbiota

By detecting specific gut microbial biomarkers in fecal samples and combining metagenomics and logistic regression models, a diabetes prediction system based on gut microbiota was constructed, which solves the problem of insufficient accuracy in diabetes diagnosis in existing technologies and achieves highly accurate non-invasive diagnosis of diabetes.

CN123120301APending Publication Date: 2026-11-17MEI YI TIAN BIOLOGICAL MEDICINE WUHAN CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611286788.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing diabetes prediction systems have limited accuracy, especially in the diagnosis of type 1 diabetes, where there is a risk of misdiagnosis. Traditional methods are insufficient to meet the diagnostic requirements of high accuracy and non-invasiveness.

Method used

By detecting specific gut microbial markers in fecal samples, such as *Fusobacterium varium*, *Fusobacterium mortiferum*, *Serratia marcescens*, *Prevotella copri*, *Megamonas funiformis*, and *Megasphaera elsdenii*, metagenomic sequencing technology was used for quantitative analysis to construct a gut microbiota-based diabetes prediction system, which was then combined with a binary logistic regression model for diagnosis.

Benefits of technology

It achieves highly accurate non-invasive diagnosis of diabetes, effectively identifies the risk of diabetes patients, especially type 1 diabetes, improves the specificity and sensitivity of the test, and fills the gap of traditional methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120301A_ABST
    Figure CN123120301A_ABST
Patent Text Reader

Abstract

This invention discloses a diabetes prediction system and model construction method based on gut microbiota, belonging to the category of disease prediction systems. It includes a detection module and a comparison module. The detection module is used to obtain quantitative detection results of preset microbial markers in the fecal samples of the test donor, including *Fusobacterium varium*. The comparison module is used to compare the quantitative detection results with preset thresholds to determine whether the test subject has diabetes based on the comparison results. It can accurately diagnose whether the test subject has diabetes and effectively assess the risk of developing diabetes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of combining biomedicine and artificial intelligence, and specifically relates to a diabetes prediction system and model construction method based on gut microbiota. Background Technology

[0002] Diabetes has become one of the most widespread chronic metabolic epidemics globally. Its harm lies not only in the continuously rising prevalence, but also in its multi-system complications affecting the heart, brain, kidneys, eyes, and peripheral nerves, as well as the lifelong dependence on exogenous insulin in type 1 diabetes (T1DM) and the potential for life-threatening acute crises such as ketoacidosis, placing a heavy burden on patients' quality of life, family finances, and social medical resources.

[0003] Data from the International Diabetes Federation (IDF) shows that approximately 537 million adults worldwide have diabetes, and the disease is showing a trend towards affecting younger people. The incidence of type 2 diabetes (T2DM) in adolescents is increasing at an alarming rate. Meanwhile, type 1 diabetes (T1DM) is not exclusive to children. Of the approximately 9.15 million patients worldwide, more than half of the new cases are adults. In China, the incidence rate among people under the age of 15 has increased nearly fourfold in the past 20 years.

[0004] Currently, the diabetes prediction systems widely used in clinical practice mainly rely on biochemical indicators such as fasting blood glucose, oral glucose tolerance test (OGTT) and glycated hemoglobin (HbA1c), and combine them with traditional risk factors such as age, body mass index and family history to construct logistic regression or scoring card models.

[0005] The aforementioned traditional diabetes prediction schemes have limited accuracy, therefore, it is necessary to improve them. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a diabetes prediction system and model construction method based on gut microbiota, which has high detection accuracy.

[0007] The technical solution provided by this invention is as follows: In a first aspect, a diabetes-related prediction system is provided, comprising: a detection module for acquiring quantitative detection results of preset microbial markers in a fecal sample of a donor to be tested, wherein the microbial markers include Fusobacterium varium; and a comparison module for comparing the quantitative detection results with preset thresholds and determining whether the person to be tested has diabetes based on the comparison results.

[0008] Secondly, a diabetes prediction system based on gut microbiota is provided, comprising: a detection module for acquiring quantitative detection results of preset microbial markers in fecal samples of a donor to be tested, wherein the microbial markers include *Fusobacterium varium*, *Fusobacterium mortiferum*, *Serratia marcescens*, *Prevotella copri*, *Megamonas funiformis*, and *Megasphaera elsdenii*; a calculation module for calculating the probability that the donor to be tested is a diabetic patient based on the quantitative detection results, denoted as a diabetic patient probability value; and a comparison module for comparing the diabetic patient probability value with a preset threshold, and determining whether the person to be tested has diabetes based on the comparison result.

[0009] Thirdly, a method for constructing a diabetes prediction model is provided. The method includes the following steps: collecting fecal samples from diabetic patients and healthy individuals according to preset inclusion and exclusion criteria, and designating the fecal samples from diabetic patients as the diabetic group and the fecal samples from healthy individuals as the healthy group; performing high-throughput sequencing on the diabetic group and the healthy group to identify differentially expressed microorganisms between the two groups, wherein the differentially expressed microorganisms include the microbial biomarkers described in the prediction system; and using the differentially expressed microorganisms as a prediction model for diabetes to determine whether the individual being tested has diabetes.

[0010] It should be noted that this invention newly discovered and verified a strong correlation between the aforementioned gut microbiota and diabetes, which can be used to diagnose whether a person being tested has diabetes. Based on this, although the embodiments of this invention only list how to achieve quantitative detection of the sample through relative abundance values, other methods for quantitative detection of microorganisms are also feasible (such as absolute abundance or total microbial load information), and can also be used to assist in the diagnosis of whether a person being tested has diabetes. Users can choose according to their own needs, and these will not be elaborated upon here.

[0011] The beneficial effects of this invention are as follows: 1. This invention newly discovers six gut microbes associated with diabetes, including: Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii.

[0012] After research and verification, it was found that for diabetic patients, one or more of the above six gut microbiota can be used to diagnose whether the person being tested has diabetes.

[0013] 2. This invention can use one or more of the above six bacterial species as detection markers (i.e., microbial markers) to diagnose whether a person has diabetes. It is completely non-invasive and highly accurate. This invention provides higher resolution through metagenomic sequencing, enabling the analysis of the microbial community to penetrate to the level of bacterial species or even strains, thereby improving the accuracy and reliability of diagnosis. The six bacterial species can also serve as target microorganisms for developing these systems, filling a gap in this field.

[0014] 3. This invention also provides a predictive system for diagnosing diabetes. This predictive system can diagnose the risk of diabetes in a test subject based on the relative abundance values ​​of various bacterial species. The entire diagnostic process is safe, non-invasive, and efficient, exhibiting good feasibility and accuracy. It is less affected by subjective response tendencies and emotional states, and can effectively assess the risk of diabetes in a test subject's (fecal sample), with relatively accurate assessment results. Attached Figure Description

[0015] Figure 1 The graph shows the results of the linear discriminant analysis on the training set. Figure 2 To validate the box plots of relative abundance in diabetic patients and healthy individuals; Figure 3 ROC curve of the predicted scores for the validation set.

[0016] Figure 4 This is a schematic diagram of the predictive system for diagnosing diabetes as described in Example 7; Figure 5 This is a schematic diagram of the structure of the gut microbiota-based diabetes prediction system described in Example 8; Figure 6 This is a flowchart illustrating the method for constructing the diabetes prediction model as described in Example 9. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to specific embodiments, so that those skilled in the art can understand it.

[0018] Diabetes has become one of the most widespread chronic metabolic epidemics globally. Its harm lies not only in the continuously rising prevalence, but also in its multi-system complications affecting the heart, brain, kidneys, eyes, and peripheral nerves, as well as the lifelong dependence on exogenous insulin in type 1 diabetes (T1DM) and the potential for life-threatening acute crises such as ketoacidosis, placing a heavy burden on patients' quality of life, family finances, and social medical resources.

[0019] Data from the International Diabetes Federation (IDF) shows that approximately 537 million adults worldwide have diabetes, and the disease is showing a trend towards affecting younger people. The incidence of type 2 diabetes (T2DM) in adolescents is increasing at an alarming rate. Meanwhile, type 1 diabetes (T1DM) is not exclusive to children. Of the approximately 9.15 million patients worldwide, more than half of the new cases are adults. In China, the incidence rate among people under the age of 15 has increased nearly fourfold in the past 20 years.

[0020] The pathogenesis of the two types of diabetes is quite different: Type 2 diabetes (T2DM) is caused by a combination of genetic susceptibility, a high-calorie diet, sedentary lifestyle, circadian rhythm disorders, and gut microbiota dysbiosis, leading to insulin resistance and progressive failure of pancreatic β-cell function; Type 1 diabetes (T1DM), on the other hand, is caused by genetic susceptibility and environmental factors triggering an autoimmune response, resulting in progressive destruction of β-cells and absolute insulin deficiency. Without timely intervention, the incidence of microvascular and macrovascular events in T2DM exceeds 50% within five years. If T1DM is diagnosed late, diabetic ketoacidosis (DKA) can be fatal, and the risk of long-term complications continues to rise. Those with poor glycemic control have significantly higher rates of disability and mortality.

[0021] For patients with diabetes (including type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM), current mainstream treatments include lifestyle interventions, oral hypoglycemic agents (metformin, SGLT-2 inhibitors, and GLP-1 receptor agonists), and exogenous insulin replacement therapy. However, long-term adherence to lifestyle interventions is less than 30%; oral medications are limited in some patients due to gastrointestinal intolerance, renal impairment, or secondary failure; while insulin is the core regimen for T1DM and advanced T2DM, it carries risks of hypoglycemia, weight gain, and poor injection adherence.

[0022] Even more serious is the long-standing neglect of the diagnostic challenges of type 1 diabetes (T1DM): it is traditionally considered a childhood disease, but 65% of cases begin in adults, and about 40% of adult T1DM patients are misdiagnosed as T2DM at initial diagnosis. In addition, a quarter of patients do not carry pancreatic autoantibodies in the early stages, and traditional classification methods have become "ineffective" due to the obesity epidemic, leading to delays in insulin treatment and greatly increasing the risk of acute crises and adverse long-term outcomes.

[0023] In response to the above difficulties, the applicant of this application filed an invention patent on November 14, 2025, entitled "Intestinal Microbial Markers, Products and Applications Related to Type 2 Diabetes (Application No.: 2025116680605)", in which the intestinal microbiota markers include any one or more of Streptococcus pharyngitis, Megacoccus ehrlich, Bifidobacterium longum, Eubacterium tumefaciens and Bifidobacterium bifidum.

[0024] While the aforementioned gut microbiota markers can be used to assess the risk of diabetes in individuals, their sample sources are primarily adults aged 18 and older, and they mainly target type 2 diabetes (T2DM).

[0025] This invention provides a microbial biomarker related to diabetes and its application. It utilizes samples from individuals of all ages, resulting in a large sample size, more accurate and representative results. This biomarker can be used as a predictor of diabetes (including T1DM and T2DM), thereby diagnosing whether a person being tested has diabetes. Specifically, the technical concept of this invention is as follows: To address the clinical needs for the diagnosis and detection of diabetes, this invention collects samples from diabetic patients (including T1DM and T2DM patients) and healthy individuals. Through standardized experimental testing procedures (specific experimental methods are described in Examples 1-3), six diabetes-related microorganisms are screened out, including: Prevotella copri (fecal), Megamonas funiformis, Megasphaera elsdenii, Fusobacterium mortiferum, Fusobacterium varium, and Serratia marcescens.

[0026] ROC curve analysis showed that the above six biomarkers have high specificity and sensitivity as detection variables, and these six bacterial species can be used as detection biomarkers for the prediction and diagnosis of diabetes patients.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. The following embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Experimental methods not specifically described in the embodiments are generally performed under conventional conditions.

[0028] Example 1: Sample Collection The inclusion criteria for the diabetes group are as follows: 1. No age limit, covering all age groups; 2. Clinically diagnosed with diabetes; 3. Stable vital signs.

[0029] Exclusion criteria for the diabetes group: 1. Received diabetes medication one month prior; 2. Took antibiotics, probiotics, or prebiotics within the past three months; 3. Suffered from any chronic disease, including neurobehavioral disorders; or had other serious chronic diseases besides diabetes, including but not limited to cardiovascular and cerebrovascular diseases, liver and kidney dysfunction, malignant tumors, autoimmune diseases, and neuropsychiatric disorders; 4. Used medications affecting gastrointestinal motility (such as prokinetic drugs, antidiarrheal drugs, etc.) within one week of enrollment; 5. History of functional dyspepsia, aerophagia, abdominal migraine, or functional constipation; 6. Showing signs of growth retardation or malnutrition; 7. Suffered from gastrointestinal obstruction or stenosis; 8. History of abdominal surgery, family history of peptic ulcers or inflammatory bowel disease; 9. Patients deemed unsuitable for inclusion in this study by the researchers.

[0030] Inclusion criteria for the control group (healthy individuals): 1. No age limit, covering all age groups; 2. No depression or other neurological diseases; 3. No diabetes or other metabolic diseases; 4. No diabetes or gastrointestinal diseases; 5. No other immune system diseases or not in an immunodeficient state; 6. No use of antibiotics (e.g., neomycin, rifaximin) or probiotics / prebiotics before and during the study.

[0031] The exclusion criteria for the control group (healthy individuals) were the same as those for the diabetes group.

[0032] The fecal samples used in this invention were collected from Hubei Province. For both diabetic patients and healthy individuals, written informed consent was obtained from both the patients and their legal guardians (if any). Based on the aforementioned inclusion and exclusion criteria, a total of 389 fecal samples were collected from diabetic patients and 438 from healthy individuals. The statistical results are as follows: Example 2: DNA extraction, library construction, and sequencing 1. Use the Hi Pure Stool DNA Mini Kit to perform DNA extraction experiments on the collected intestinal samples.

[0033] 2. After extraction, the DNA concentration was detected using Qubit, and the integrity of the extracted genomic DNA was detected by 1.5% agarose gel electrophoresis. The extracted genomic DNA was then subjected to quality control, and qualified genomic DNA samples were selected (DNA concentration ≥20ng / μL, volume ≥20 μL, total amount ≥400 ng).

[0034] 3. For qualified DNA samples, random fragmentation, end repair, A base ligation, adapter and index are added. After adapter ligation, purification and library amplification are performed. After amplification, the DNA concentration is detected (DNA concentration ≥40 ng / μL).

[0035] 4. After the libraries pass the testing, different libraries are pooled according to the effective concentration and target data volume requirements before sequencing. The metagenomic sequencing platform is BGI T7, and the sequencing strategy is PE150.

[0036] Example 3: LEfSe analysis for screening microbial biomarkers 1. Split the dataset KneadData software was used for quality control (based on Trimmomatic) and host removal (based on Bowtie2) of the raw data. Kraken2 alignment was used to calculate the sequence number of each species in the sample, and Bracken was used to estimate the actual abundance of species in the sample. 80% of the participants (including those in the disease and healthy groups) were randomly selected as the training set, and the remaining 20% ​​of the samples were used as the validation set. The abundance data of each sample in the training set were then analyzed using LEfSe software, with the default LDA Score filter value set to 2. The sample information is shown in Table 1. The results are as follows Figure 1 As shown, the researchers screened out six biomarkers that were significantly increased in the diabetes group, including fecal Prevotella copri, Megamonas funiformis, Megasphaera elsdenii, Fusobacterium mortiferum, Fusobacterium varium, and Serratia marcescens.

[0037] Example 4: Verifying the reliability of the above 6 microbial biomarkers 1. First, the remaining 20% ​​of the participants in Example 1 (including those in the diabetes group and the healthy group) were used as the validation set. The abundance data of each sample in the validation set were first subjected to binary logistic regression, and then the receiver operating characteristic (ROC curve) was analyzed to obtain the cutoff value (optimal cutoff value).

[0038] 2. Use R language statistical software to calculate specificity and sensitivity and plot ROC curves. The software first calculates the threshold of the actual measurement value, and then calculates the number of true positive cases (TP), false positive cases (FP), true negative cases (TN), and false negative cases (FN) corresponding to the threshold. Specificity (true negative rate) = TN / (TN + FP) Sensitivity (true positive rate) = TP / (TP + FN) 3. The ROC curve can be constructed using 1 minus specificity and sensitivity. The integral of the ROC curve is the AUC. To calculate the specificity and sensitivity of a certain indicator, first calculate the Youden coefficient (Youden index = sensitivity + specificity - 1). The specificity and sensitivity corresponding to the maximum value of the Youden coefficient are the specificity and sensitivity of the certain indicator.

[0039] 4. The relative abundance values ​​of microbial biomarkers for single strains were directly analyzed using receiver operating characteristic (ROC) curve testing to determine the cutoff value. The ROC curve for predictive scoring is shown below. Figure 3 As shown in Table 2, the AUC, optimal cutoff value, sensitivity, and specificity of the predicted mimicry markers (markers formed by the combination of 6 single bacterial species) and individual bacteria are shown in Table 2.

[0040] Table 2 ROC diagnostic curve results As shown in Table 2, for single bacterial species, *Megacoccus escherichiae* had the highest AUC value (approximately 0.812), while *Megamonas filamentosa* had the lowest AUC value (approximately 0.729). The AUC value of the mimicry marker (a marker formed by the combination of 6 single bacterial species) (approximately 0.984) was higher than that of the single bacterial species.

[0041] As can be seen from the above, one or more of the six newly discovered bacterial species in this application can be used as detection variables, and they all have high specificity and sensitivity. Moreover, the AUC of all six microbial markers is greater than 70%. Therefore, one or more of the six microbial markers can be used as detection markers for the diagnosis of diabetic patients.

[0042] Example 5: Establishing a Logistic Regression Model Based on the microbial biomarkers selected above and the relative abundance value of each metabolic biomarker, the binary logistic regression algorithm in RStudio software was used to calculate the first disease probability (also known as the logarithm of dominance y) for each sample. Then, the disease probability optimization formula Z=exp(y) / {1+exp(y)} was used to calculate the disease probability Z of the sample under test. Finally, this disease probability Z was compared with the actual disease status (e.g., severity) of each sample to verify the accuracy of the disease probability calculation equation. Specifically: a. Establishing a model Based on the biomarkers identified above and the proportion of diabetic patients in the training set, the relative abundance values ​​of the six detected bacterial species were further used as single variables. The linear relationship between the relative abundance values ​​of the six single bacteria and the disease probability of the samples was then discussed. The logarithm y (also known as the first probability value y) of the dominance of the test subject was calculated using a binary logistic regression equation. y=A+B1×x1+B2×x2+B3×x3+B4×x4+B5×x5+B6×x6; Where A is the intercept term, B1 to B6 are the regression coefficients of the independent variables; x1 is the relative abundance value of Prevotella copri in feces, x2 is the relative abundance value of Megamonas funiformis, x3 is the relative abundance value of Megasphaera elsdenii, x4 is the relative abundance value of Fusobacterium mortiferum, x5 is the relative abundance value of Fusobacterium varium, and x6 is the relative abundance value of Serratia marcescens.

[0043] b. Determine the values ​​of A and B1 to B6 above. After statistical analysis of the sample data, A was -0.65033, B1 was 9.80499, B2 was 55.70690, B3 was 54.56902, B4 was 472.47182, B5 was 638.23985, and B6 was 1099.69308. At this point, after rearrangement, the formula for calculating the logarithm y of the dominance is: y=-0.65033+9.80499×X1+55.70690×X2+54.56902×X3+472.47182×X4+638.23985×X5+1099.69308×X6; c. Calculate the disease probability Z of the subjects to be tested. Substitute the first probability value y into the following formula to calculate the probability Z that the subject is a patient: Z = exp(y) / {1 + exp(y)}; where Z is the probability value that the subject is a patient, and exp(y) is the natural exponential function of the first probability value y.

[0044] After processing, the formula for calculating the probability Z of disease is: ; d. Validation set data calculation and statistical analysis Based on the validation set data, the relative abundance of each single bacterial species in the disease group and the healthy group for each sample was obtained. Then, the first probability value y was obtained using the aforementioned binary logistic regression method. The probability Z of the test sample being a patient was then calculated using the formula. The results are shown in Tables 3 and 4, where "patient" refers to a patient with irritable bowel syndrome.

[0045] Table 4 shows the relative abundance mean and standard deviation for each bacterial species. The relative abundance mean determines the central location of the data distribution, while the standard deviation reflects the dispersion of the data relative to the mean. The p-value is a statistic calculated using the rank-sum test formula. The lower the p-value, the greater the difference between the disease group and the healthy group.

[0046] Table 3. Relevant data on validation set markers Table 4. Relevant data on validation set markers Note: e is a power of 10, for example, express Note: In Table 4, the mean refers to the relative abundance mean, and the standard deviation is similarly calculated.

[0047] In the table above, composition 1 consists of *Prevotella feces* + *Megalomonas cordifolia* + *Megacoccus escherichiae* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Serratia depigmentosus*; composition 2 consists of *Prevotella feces* + *Megalomonas cordifolia* + *Megacoccus escherichiae* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis*; composition 3 consists of *Megalomonas cordifolia* + *Megacoccus escherichiae* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Serratia depigmentosus*; composition 4 consists of *Prevotella feces* + *Megalomonas cordifolia* + *Megacoccus escherichiae* + *Fusobacterium catarrhalis*; composition 5 consists of *Megacoccus escherichiae* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Serratia depigmentosus*; composition 6 consists of *Prevotella feces* + *Megalomonas cordifolia* + *Megacoccus escherichiae*; composition 7 consists of *Fusobacterium catarrhalis* + *Fusobacterium catarrhalis* + *Serratia depigmentosus*; and composition 8 consists of *Prevotella feces* + *Megalomonas cordifolia*.

[0048] e. Results and Analysis Based on the results of Examples 4 and 5, it can be seen that, for single bacterial species, *Megacoccus ehrlich* had the highest AUC value (approximately 0.812), while *Megamonas cordifolia* had the lowest AUC value (approximately 0.729). The AUC value of the mimicry marker (a marker formed by the combination of 6 single bacterial species) (approximately 0.984) was higher than that of the single bacterial species.

[0049] The AUC of the six microbial markers identified in this application is all greater than 70%. One or more of the six microbial markers can be used as detection markers for the diagnosis of diabetic patients. At the same time, since this invention only requires the collection of stool samples from the test subjects, this invention is not only completely non-invasive, but also highly accurate, and can be used to diagnose whether the test subjects (stool samples) have diabetes.

[0050] Based on the results obtained from the table above and the calculation formula for the probability of disease Z, it can be seen that the calculation formula for calculating the probability of disease in the sample to be tested, which is summarized by this application based on the relative abundance values ​​of the six microbial markers, is basically correct and can be used to diagnose the risk and probability of disease in the sample to be tested. The probability of disease in the table above may not fully meet the diagnostic criteria. This is because the intestinal samples of the person to be tested may produce false positive or false negative results. Further testing using other methods is required, including blood routine tests, diagnostic physical signs, etc.

[0051] Example 6: Examining the probability of disease in patients in a validation set and diabetic patients Based on the product and method of Example 5, the disease probability of healthy individuals and diabetic patients in the verification group was examined and verified. The specific steps are as follows: S1. Collect intestinal samples from the individuals to be tested and detect the relative abundance of each individual strain in the intestine; wherein, the individual strains include Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii.

[0052] S2. Calculate the logarithm y of the strength of the object under test based on the binary logistic regression equation; y=A+B1×x1+B2×x2+B3×x3+B4×x4+B5×x5+B6×x6; Where A is the intercept term, B1 to B6 are the regression coefficients of the independent variables; x1 is the relative abundance value of Prevotella copri in feces, x2 is the relative abundance value of Megamonas funiformis, x3 is the relative abundance value of Megasphaera elsdenii, x4 is the relative abundance value of Fusobacterium mortiferum, x5 is the relative abundance value of Fusobacterium varium, and x6 is the relative abundance value of Serratia marcescens.

[0053] S3. Calculate the probability Z of the subject being a diabetic patient based on y, Z = exp(y) / {1 + exp(y)}; exp(y) is the natural exponential function of y; The probability of disease Z can also be expressed as: ; S4. Based on the comparison between the patient's probability Z-value and the reference value, diagnose or predict the risk of the subject having diabetes.

[0054] In practice, a Z-score greater than 0.5 indicates a higher probability that the subject has diabetes; a Z-score less than 0.5 indicates a lower probability; and a Z-score of 0.5 suggests the subject may be a patient or have diabetes, requiring further testing using other methods such as mental health assessments and medication testing. Furthermore, the closer the Z-score is to 0.5, the more necessary it is to utilize additional testing methods.

[0055] It should be noted that although this embodiment only lists the methods and means of quantitative detection of the sample to be tested through relative abundance value, other means of quantitative detection of microorganisms are also feasible for the present invention (such as absolute abundance or total microbial load information, etc.), and can also be used to assist in the diagnosis of whether the sample to be tested has diabetes. People can choose the appropriate quantitative detection method of microorganisms according to their own needs, which will not be elaborated here.

[0056] Example 7: Predictive system for diagnosing diabetes Based on the above embodiments, this embodiment provides a diabetes-related prediction system for assessing the risk of a subject being a diabetic patient. The prediction system includes a detection module and a comparison module, specifically: The detection module is used to obtain quantitative detection results of a single bacterial species in the fecal sample of the test subject; wherein, the single bacterial species may include one or more of the following: Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii; The comparison module is used to compare the quantitative detection results with a preset threshold, and to determine the risk of the subject being a diabetic patient based on the comparison results.

[0057] Preferably, the quantitative detection results include one or more of relative abundance, absolute abundance, and total microbial load information, and the preset threshold is a range of values ​​determined by the mean and standard deviation of the quantitative detection results.

[0058] In practical application, to illustrate how the prediction system assesses the risk of a test subject having diabetes, this invention uses relative abundance values ​​as an example. Specifically: The detection module obtains the relative abundance value of Fusobacterium varium in the fecal sample of the test subject, which is recorded as the first abundance value. This first abundance value is also the quantitative detection result obtained by the detection module. The comparison module compares the first abundance value with the mean and standard deviation of Fusobacterium varium in Table 4 to assess the risk that the subject is a diabetic patient. For example, if the first abundance value is within the range determined by the mean and standard deviation of diabetic patients, the risk of the subject being a diabetic patient is high, and vice versa.

[0059] Similarly, referring to the detection and assessment method for Fusobacterium varium mentioned above, the prediction system can also perform similar detection and assessment on one or more of Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii, thereby assessing the risk of the test subject having diabetes.

[0060] Due to factors such as genetics and environment, it is normal that not all six newly discovered bacterial species can be detected in fecal samples. People can use one or more of the six newly discovered bacterial species as detection markers to assess the risk of the subject being a diabetic patient.

[0061] Given that obtaining relative abundance values ​​and other quantitative detection data (such as absolute abundance or total microbial load information) are routine methods in this field, if one wants to use other quantitative detection methods to assess the risk of a subject being a diabetic patient, one can refer to the description above, which will not be repeated here.

[0062] Example 8: Diabetes Prediction System Based on Gut Microbiota Based on the above embodiments, this embodiment provides a diabetes diagnosis / prediction system, including a detection module, a calculation module, and a comparison module, wherein: The detection module is used to obtain quantitative detection results of preset microbial markers in the fecal samples of the donor to be tested. The microbial markers include Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii.

[0063] The calculation module is used to calculate the probability that the donor to be tested is a diabetic patient based on the quantitative detection results, and denoted as the diabetic patient probability value.

[0064] In actual operation, the calculation module also includes a first probability module and a second probability module, wherein: the first probability module is used to substitute the relative abundance value of each single species in the microbial biomarker into the binary logistic regression equation to calculate the logarithm y of the dominance of the test subject; the second probability module is used to calculate the probability Z of the test subject being a type 2 diabetic patient based on y, Z=exp(y) / {1+exp(y)}, where exp(y) is an exponential function of y.

[0065] The comparison module is used to compare the probability value of the diabetic patient with a preset threshold, and to determine whether the person being tested has diabetes based on the comparison result.

[0066] Example 9: Method for constructing a diabetes prediction model Based on the descriptions in Examples 1-7, the present invention provides a method for constructing a prediction model for diabetes, comprising the following steps: S1. Sample collection: Collect fecal samples from diabetic patients and healthy individuals according to the preset inclusion and exclusion criteria. In this step, given that different groups of people have different definitions of "healthy person," such as some people considering family members of long-lived families as healthy people (passing physical examinations), some people considering young people aged 20-30 as healthy people (passing physical examinations), some people considering college students of sports colleges as healthy people (passing physical examinations), etc., people can use the healthy person screening method listed in Example 1 of this invention to screen for healthy people, or they can use other standards to screen for healthy people. This embodiment does not impose any restrictions here, as long as the collected fecal samples conform to people's general understanding.

[0067] S2. Sample grouping: DNA extraction, library construction and sequencing were performed on the fecal samples of the above-mentioned diabetic patients and healthy individuals respectively. The fecal samples of diabetic patients were assigned to the diabetic group and the fecal samples of healthy individuals were assigned to the healthy group. In this step, after selecting specific fecal samples from healthy individuals, people should be able to extract DNA from the fecal samples, construct libraries, and sequence them according to the preset steps to distinguish the differential microorganisms between the diabetic group and the healthy group. The specific experimental steps can be referred to the descriptions in Examples 1 to 4 above, and will not be repeated here.

[0068] S3. Identify differentially expressed microorganisms: Perform high-throughput sequencing on the diabetic group and the healthy group to identify differentially expressed microorganisms between the diabetic group and the healthy group. The differentially expressed microorganisms may include one or more of the following: Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii. The steps for performing high-throughput sequencing and identifying differentially expressed microorganisms can be found in the descriptions of Examples 1-4 above, and will not be repeated here.

[0069] S4. Use the differentially expressed microorganisms as a predictive model for diabetes to distinguish whether the person being tested is a diabetic patient.

[0070] In practical work, in conjunction with Examples 7 and 8, this step provides two methods for using the differentially expressed microorganisms as predictive models, specifically including: Referring to Example 7, the first method of using the differential microorganisms as a prediction model is as follows: S41, obtain the quantitative detection results of preset microbial markers in the fecal sample of the person to be tested; S42, compare the quantitative detection results with preset thresholds, and determine whether the person to be tested is a diabetic patient based on the comparison results.

[0071] Referring to Example 8, the second method of using the differential microorganisms as a prediction model is as follows: S41, obtain the quantitative detection results of preset microbial markers in the fecal sample of the person to be tested; S42, calculate the probability that the person to be tested is a diabetic patient based on the quantitative detection results, and use the probability value of the diabetic patient as the quantitative detection result; S43, compare the quantitative detection result with a preset threshold, and determine whether the person to be tested is a diabetic patient based on the comparison result.

[0072] In actual work, the above steps S41 to S43 can also be the following steps: K41. Obtain the relative abundance value of each single bacterial species in the feces of the test subject; the single bacterial species include Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii; K42. Substitute the relative abundance value of each single bacterial species into the binary logistic regression equation to calculate the logarithm y of the dominance of the test object (i.e., the first probability value y). y=A+B1×x1+B2×x2+B3×x3+B4×x4+B5×x5+B6×x6; Where A is the intercept term, B1 to B6 are the regression coefficients of the independent variables; x1 is the relative abundance value of Prevotella copri in feces, x2 is the relative abundance value of Megamonas funiformis, x3 is the relative abundance value of Megasphaera elsdenii, x4 is the relative abundance value of Fusobacterium mortiferum, x5 is the relative abundance value of Fusobacterium varium, and x6 is the relative abundance value of Serratia marcescens.

[0073] K43. Calculate the probability Z of the subject being a diabetic patient based on y, Z = exp(y) / {1 + exp(y)}; exp(y) is the natural exponential function of y; The probability of disease Z can also be expressed as: .

[0074] K44. Based on the comparison between the patient's probability Z-score and the reference value, diagnose or predict the risk of the subject having diabetes.

[0075] In practice, a Z-score greater than 0.5 indicates a higher probability that the subject has diabetes; a Z-score less than 0.5 indicates a lower probability; and a Z-score of 0.5 suggests the subject may be a patient or have diabetes, requiring further testing using methods such as mental health assessments and medication testing. Furthermore, the closer the Z-score is to 0.5, the more necessary it is to utilize additional testing methods.

[0076] Example 10: Diagnosing whether a test sample belongs to a diabetic patient If it is necessary to diagnose whether a person being tested has diabetes, in addition to conventional testing methods such as ultrasound examination, serum liver enzyme test, and liver tissue biopsy, medical staff can also use the methods or products described in Examples 1-9 above to diagnose the person being tested, in order to assist medical staff in making a more accurate judgment: (1) If, after continuous observation over multiple time periods, the content of Prevotella copri, Megamonas funiformis, Megasphaera elsdenii, Fusobacterium mortiferum, Fusobacterium varium, and Serratia marcescens in the feces of the test subject is found to be high (compared to the mean and standard deviation of the diabetes group in Table 4), or even shows a significant increasing trend, then the test subject is more likely to be a diabetic patient.

[0077] (2) If, after observation over multiple consecutive periods, the person being tested does not exhibit the pattern described in (1) above, then medical staff can combine [the above information with further details]. Figure 1 According to Table 4, a preliminary judgment is made on whether the person being tested is a diabetic patient; (3) If medical staff want to more accurately determine whether the person to be tested is a diabetic patient, they can calculate the probability that the person to be tested is a diabetic patient based on the quantitative detection results of 6 single bacterial species (such as relative abundance values) and with reference to the technical solutions described in Examples 5 to 7.

[0078] Conclusion and explanation: 1. By Figures 1-3 As shown in Table 2, any one of the six newly discovered single bacterial species in this application can serve as a microbial biomarker for diabetes. Each single bacterial species exhibits both sensitivity and specificity for diabetes. Therefore, the microbial biomarker for diabetes can be selected from any one or more of the six newly discovered single bacterial species in this application. Specifically: Microbial markers for diabetes can be selected from one or more of the following: Prevotella copri, Megamonas funiformis, Megasphaera elsdenii, Fusobacterium mortiferum, Fusobacterium varium, and Serratia marcescens.

[0079] 2. As shown in Table 3, it is normal for only one or a few species of bacteria to be detected when testing intestinal samples. This is because there are individual differences. The probability of disease Z in this application is calculated. Therefore, even if a sample contains only a single species of bacteria, this application can still calculate the probability that the sample has diabetes.

[0080] 3. Predictive effect: The AUC value of the mimicry marker (a marker formed by the combination of 6 single bacterial species) (approximately 0.983) is higher than that of a single bacterial species. When other microorganisms are added to one or several single bacterial species for testing, the test result (AUC value) will not be lower than the lowest AUC value of multiple single bacterial species.

[0081] 4. All six newly discovered bacterial species in this application can be used as detection variables. They all have high specificity and sensitivity, and the AUC of all six microbial markers is greater than 75%. One or more of the six microbial markers can be used as detection markers for the diagnosis of diabetic patients.

[0082] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the invention.

Claims

1. A predictive system for diagnosing diabetes, characterized in that, include: The detection module is used to obtain quantitative detection results of preset microbial markers in the fecal sample of the donor to be tested, wherein the microbial markers include Fusobacterium varium; The comparison module is used to compare the quantitative detection results with a preset threshold and determine whether the person being tested has diabetes based on the comparison results.

2. The prediction system according to claim 1, characterized in that, The microbial markers also include *Fusobacterium mortiferum* and / or *Serratia marcescens*.

3. The prediction system according to claim 1 or 2, characterized in that, The microbial markers also include one or more of the following: Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii.

4. A diabetes prediction system based on gut microbiota, characterized in that: include: The detection module is used to obtain quantitative detection results of preset microbial markers in the fecal samples of the donor to be tested. The microbial markers include Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii. The calculation module is used to calculate the probability that the donor to be tested is a diabetic patient based on the quantitative detection results, and denoted as the diabetic patient probability value; The comparison module is used to compare the probability value of the diabetic patient with a preset threshold, and determine whether the person being tested has diabetes based on the comparison result.

5. The diabetes prediction system according to claim 4, characterized in that: The calculation module further includes a first probability module and a second probability module, wherein: The first probability module is used to substitute the relative abundance value of each single species in the microbial biomarker into the binary logistic regression equation to calculate the logarithm y of the dominance of the test object; The second probability module is used to calculate the probability Z of the subject being a diabetic patient based on the test y, Z = exp(y) / {1 + exp(y)}, where exp(y) is an exponential function of y.

6. The prediction system according to claim 5, characterized in that: The formula for the binary logistic regression equation is: y=A+B1×x1+B2×x2+B3×x3+B4×x4+B5×x5+B6×x6; Where A is the intercept term, B1 to B6 are the regression coefficients of the independent variables; x1 is the relative abundance value of Prevotella copri in feces, x2 is the relative abundance value of Megamonas funiformis, x3 is the relative abundance value of Megasphaera elsdenii, x4 is the relative abundance value of Fusobacterium mortiferum, x5 is the relative abundance value of Fusobacterium varium, and x6 is the relative abundance value of Serratia marcescens.

7. The prediction system according to claim 6, characterized in that: The values ​​are: A = -0.65033, B1 = 9.80499, B2 = 55.70690, B3 = 54.56902, B4 = 472.47182, B5 = 638.23985, and B6 = 1099.69308.

8. A method for constructing a diabetes prediction model, characterized in that, The method for constructing the prediction model includes the following steps: Fecal samples were collected from diabetic patients and healthy individuals according to the pre-set inclusion and exclusion criteria. The fecal samples from diabetic patients were designated as the diabetic group, and the fecal samples from healthy individuals were designated as the healthy group. High-throughput sequencing was performed on the diabetic group and the healthy group to identify differentially expressed microorganisms between the diabetic group and the healthy group, wherein the differentially expressed microorganisms include the microbial biomarkers described in the prediction system of any one of claims 1 to 7; The differentially expressed microorganisms were used as a predictive model for diabetes to determine whether the individuals being tested had diabetes.

9. The construction method according to claim 8, characterized in that, The use of the differentially expressed microorganisms as a predictive model for diabetes includes: Obtain quantitative detection results of pre-defined microbial markers in fecal samples from the donor to be tested; The quantitative detection results are compared with a preset threshold, and the donor to be tested is determined to be either a diabetic patient or an inefficient donor based on the comparison results.

10. The construction method according to claim 8, characterized in that, The use of the differentially expressed microorganisms as a predictive model for diabetes includes: Quantitative detection results of preset microbial markers in fecal samples from the donor to be tested are obtained. The microbial markers include Fusobacterium varium, Fusobacterium mortiferum, Serratia marcescens, Prevotella copri, Megamonas funiformis, and Megasphaera elsdenii. The probability that the donor to be tested is a diabetic patient is calculated based on the quantitative test results, and the probability value of the diabetic patient is used as the quantitative test result. The quantitative detection results are compared with a preset threshold, and the test subject is judged to determine whether he or she has diabetes based on the comparison results.