Microbial marker combination for diagnosing diabetes after kidney transplantation and application thereof

Abstract

The application discloses a microorganism marker combination for diagnosing diabetes after kidney transplantation and application thereof. The microorganism marker combination takes Succinivibrio, Akkermansia, Desulfoarcina and Schwartziella as core diagnostic markers, and the area under the ROC curve of a combined diagnostic model constructed from the core diagnostic markers reaches 0.981, and the diagnostic sensitivity and specificity are significantly better than those of a single marker. High-precision early diagnosis of PTDM is achieved, and a non-invasive, convenient and efficient new means is provided for screening and intervention of PTDM.

Description

Technical Field

[0001] This invention relates to the field of biomedical detection technology, specifically to a combination of microbial biomarkers for diagnosing diabetes after kidney transplantation and their application. Background Technology

[0002] Chronic kidney disease (CKD) is a major global public health problem. Kidney transplantation is the best treatment option for end-stage renal disease, but kidney transplant recipients are prone to various complications after the procedure. Among them, the incidence of new-onset diabetes mellitus (PTDM) after transplantation is as high as 10% to 40%, which is a key complication affecting the long-term survival rate of patients and the survival of transplanted kidneys.

[0003] Persistent PTDM is characterized by insufficient insulin secretion and insulin resistance, and its clinical manifestations are similar to those of type 2 diabetes. However, its pathogenesis is influenced by a triple factor: chronic kidney disease, immunosuppressant use, and metabolic disorders, and its pathogenesis remains unclear. Currently, clinical diagnosis of PTDM mainly relies on hematological indicators such as blood glucose and glycated hemoglobin. These indicators only reflect abnormalities in glucose metabolism and cannot provide early prediction; furthermore, adjusting immunosuppressive regimens to intervene in PTDM carries the risk of graft rejection, making these tests invasive. Therefore, there is an urgent clinical need for non-invasive, early, and accurate diagnostic methods for PTDM.

[0004] Microbiome studies have shown that the oral microbiota is a sensitive indicator reflecting the host's metabolic state and immune function, and that oral microbiota dysbiosis is closely related to abnormal glucose metabolism and chronic kidney disease. The oral microecology of kidney transplant recipients is specifically altered due to disease and immunosuppressive therapy, and periodontal pathogens can participate in the development of post-transplant complications through immune modulation. However, oral microbiome research targeting specific clinical scenarios of PTDM is still in its early stages. There are no studies systematically revealing the characteristics of the oral microbiome composition of PTDM patients, and there is also a lack of early diagnostic tools for PTDM based on the oral microbiome.

[0005] Based on the above problems, this invention systematically analyzes the differences in oral microbial composition and function between PTDM patients and normal glucose tolerance after kidney transplantation, screens out oral microbial biomarkers with high diagnostic value, and constructs an efficient joint diagnostic model, realizing non-invasive and accurate diagnosis of PTDM, filling the technological gap in this field. Summary of the Invention

[0006] The purpose of this invention is to provide a combination of microbial biomarkers for diagnosing post-renal transplant diabetes mellitus (PTDM) and its application; it also provides a method for constructing a PTDM prediction model based on the oral microbiome. This solves the problems of existing PTDM diagnostic methods being unable to predict early and requiring invasive testing, providing a non-invasive, convenient, and highly accurate technical solution for clinical PTDM screening and early intervention.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A microbial biomarker combination for diagnosing diabetes after kidney transplantation comprises four core oral microbial biomarkers: *Succinivibrio*, *Akkermansia*, *Succiniclasticum*, and *Schwartzia*. The biomarkers are relative abundance at the genus level, and the test sample is saliva from the oral cavity of a kidney transplant recipient.

[0008] A diagnostic kit comprising reagents for detecting the combination of the above-mentioned microbial markers, said reagents comprising: (1) Saliva microbial DNA extraction reagent; (2) Primers for amplification of the 16S rDNA V3-V4 region: 341F: 5'-CCTACGGGNGGCWGCAG-3', 805R: 5'-GACTACHVGGGTATCTAATCC-3'; (3) PCR amplification reagents, purification reagents, and sequencing library construction reagents; (4) Negative control, positive control and abundance quantification reference.

[0009] The application of the microbial biomarker combination or the kit described in this invention in the preparation of a non-invasive diagnostic product for new-onset diabetes mellitus after kidney transplantation. The diagnostic product is used to: (1) detect the 16S rDNA V3-V4 region sequence of saliva microorganisms; (2) calculate the relative abundance of four core microbial genera; and (3) input the results into a combined diagnostic model to output the positive / negative results for PTDM.

[0010] The genomic DNA extraction reagents and abundance detection reagents described in this invention are well known to those skilled in the art and can be obtained through conventional means, and therefore are not particularly limited.

[0011] A method for constructing a predictive model for diabetes after kidney transplantation based on the oral microbiome includes the following steps: (1) Sample grouping: Oral saliva was collected from subjects ≥3 months after kidney transplantation and divided into a PTDM case group and a normal glucose tolerance control group, excluding individuals with preoperative diabetes and stress hyperglycemia; (2) DNA extraction and PCR amplification: Microbial DNA was extracted and the 16S rDNA V3-V4 region was amplified using primers 341F / 805R. (3) Sequencing and noise reduction: NovaSeq6000 paired-end sequencing, DADA2 noise reduction, ASV clustering, SILVA database annotation, and microbial abundance matrix were obtained. (4) Biomarker screening: Mann-Whitney U difference test, LEfSe analysis and random forest model were used to screen differentially expressed species, and the intersection was taken to obtain the core oral microbial biomarkers; (5) Model building: A binary logistic regression joint diagnostic model was constructed based on the relative abundance of four core oral microbial biomarkers. (6) Performance evaluation: ROC curves were generated and AUC was calculated to determine the diagnostic threshold, and the sensitivity, specificity, and accuracy of the model were verified to evaluate its predictive efficacy. The diagnostic threshold is -0.99. A Y value ≥ -0.99 is considered positive for PTDM, and a Y value < -0.99 is considered negative. The model has an AUC ≥ 0.981, sensitivity ≥ 100%, and specificity ≥ 96.6%.

[0012] Before collecting oral samples in step (1) of this invention, the subject fasted for 1 hour from food, water, alcohol, tobacco and chewing gum. Saliva samples were collected from the bilateral gingival margins and buccal mucosa and stored at -80°C for 2 hours.

[0013] The primer sequences for PCR amplification described in this invention are as follows: 341F: 5'-CCTACGGGNGGCWGCAG-3', 805R: 5'-GACTACHVGGGTATCTAATCC-3'; The PCR amplification program was as follows: denaturation at 98℃ for 1 min, followed by 35 cycles of denaturation at 98℃ for 10 s, annealing at 54℃ for 30 s, extension at 72℃ for 45 s, and extension at 72℃ for 10 min.

[0014] The core oral microbial markers mentioned in step (4) of this invention are Vibrio succinate, Akkermania succinate, Vibrio schwarzia succinate, and Vibrio schwarzia succinate.

[0015] The equation of the binary logistic regression joint diagnostic model described in this invention is: Y = -1.504 + 0.006 × Akkermania spp. relative abundance + 0.681 × Vibrio succinate relative abundance - 0.521 × Vibrio succinate relative abundance + 9.072 × Schwarzia spp. relative abundance. The combined model has an AUC of not less than 0.981, which is significantly higher than the diagnostic efficacy of a single microbial biomarker.

[0016] Beneficial effects: 1. This invention uses oral saliva as the test sample. The collection process is non-invasive, simple to operate, and has high subject compliance, making it suitable for long-term screening of kidney transplant recipients.

[0017] 2. This invention provides accurate diagnosis. The combined diagnostic model constructed from the core biomarker combination achieves an AUC of 0.981, demonstrating extremely high diagnostic efficacy and enabling accurate diagnosis of PTDM. The auxiliary biomarker combination can further expand the diagnostic system to meet different clinical testing needs.

[0018] 3. This invention allows for early prediction. Changes in the oral microbiota precede abnormalities in hematological indicators such as blood glucose. This invention enables early prediction of PTDM, providing a time window for early clinical intervention.

[0019] 4. The invention has high specificity. The biomarkers of this invention are obtained based on a large-sample screening of kidney transplant recipients and are applicable to kidney transplant recipients receiving a triple immunosuppressive regimen of tacrolimus + mycophenolic acids + glucocorticoids, exhibiting high specificity for clinical application. Attached Figure Description

[0020] Figure 1 Oral microbiome sequencing analysis of recipients in the PTDM group and control group (A: Box plot of high-quality sequences in the two groups, B: Venn plot of ASV count in the two groups) Figure 2 : Oral microbial β-diversity analysis diagrams of PTDM group and control group, where A is PCoA analysis diagram and B is NMDS analysis diagram; Figure 3 ROC curve analysis of each individual marker in the core marker combination. Detailed Implementation

[0021] The technical solution of the present invention will be further described in detail below through specific embodiments.

[0022] Implementation 1: Screening and Model Building of Microbial Biomarker Combinations for the Diagnosis of Diabetes After Kidney Transplantation 1. Sample collection 1.1 Study subjects: 61 kidney transplant recipients from the Department of Organ Transplantation, Affiliated Hospital of Guizhou Medical University were included, including 30 patients with PTDM and 31 patients with normal glucose tolerance after surgery (control group).

[0023] Inclusion criteria: age > 18 years; more than 3 months post-kidney transplant; receiving a triple immunosuppressive regimen of tacrolimus + mycophenolate mofetil + glucocorticoids; and stable transplanted kidney function.

[0024] Exclusion criteria: combined organ transplantation; poor compliance; graft rejection; preoperative history of diabetes; transient hyperglycemia caused by stress or high doses of corticosteroids.

[0025] Diagnostic criteria for PTDM: Following the standards of the American Diabetes Association and the World Health Organization, meeting any of the following conditions: fasting blood glucose >7 mmol / L; random blood glucose >11.1 mmol / L with symptoms of diabetes; 2-hour blood glucose ≥11.1 mmol / L after a 75g oral glucose tolerance test; HbA1c >6.5%. Based on these criteria, recipients diagnosed with PTDM were included in the PTDM group, and recipients with normal glucose tolerance were included in the normal control group.

[0026] 1.2 Sample collection: All subjects fasted for 1 hour, refrained from drinking water, alcohol, smoking, and chewing gum. The sampler wore sterile gloves and gently scraped the bilateral gingival margins (10 times) and buccal mucosa (10 times) with a sterile pharyngeal swab. The sampling end was placed into a sterile EP tube along the sterile break point of the swab handle and placed in a low-temperature incubator. The tube was then sent to the laboratory for freezing at -80°C within 2 hours.

[0027] 2. Sequencing of the 16S rDNA V3-V4 region of microorganisms in saliva samples 2.1 Extraction and quality control of DNA from saliva samples (1) After thawing the collected saliva, transfer it to a 2ml sterile centrifuge tube; (2) Add an appropriate amount of cetyltrimethylammonium bromide extraction buffer (about 700 µL) to the centrifuge tube, gently invert to mix, and ensure uniform dispersion; (3) Incubate the mixture in a 65°C water bath for 30 minutes to promote cell lysis and protein denaturation. Then, centrifuge (12,000 rpm, 10 minutes) and collect the supernatant; (4) Add an equal volume of chloroform to the supernatant, gently invert to mix, then centrifuge (12000 rpm, 10 minutes) and collect the supernatant; (5) Add 2.5 times the volume of 70% ethanol to the supernatant, gently invert to mix, and then place in a -20°C refrigerator for 30 minutes to precipitate DNA. Then centrifuge the sample (12000 rpm, 10 minutes) and discard the supernatant; (6) Wash the precipitated DNA with 70% ethanol, gently invert to mix, then remove the ethanol and dry the DNA precipitate. Finally, resuspend the DNA precipitate in TE buffer; (7) Mix DNA and loading buffer with 2% agarose gel, load the sample, perform electrophoresis, stain with dye, and then perform quality assessment.

[0028] 2.2 PCR amplification and purification (1) PCR amplification: The extracted DNA was used as a template for PCR. Primer pairs 341F (5'-CCTACGGGNGGCWGCAG-3') and 805RR (5'-GACTACHVGGGTATCTAATCC-3') were used to amplify and analyze the 16S rDNA V3-V4 region. Specific steps are as follows: First, the reaction system was denatured at 98°C for 1 minute. Then, 35 cycles of amplification were performed: in each cycle, denaturation was first performed at 98°C for 10 seconds, followed by annealing at 54°C for 30 seconds, and finally extension at 72°C for 45 seconds. After amplification, the extension was maintained at 72°C for 10 minutes. Finally, the sample should be stored at 4°C for subsequent analysis.

[0029] (2) PCR purification and quantification: The PCR amplification products were purified using AMPure XP beads to remove excess primers, enzymes, and other reaction components. After purification, the concentration of the PCR products was accurately measured using the Qubit quantitative PCR method (Invitrogen, USA).

[0030] (3) PCR product recovery: The amplified PCR products were detected by 2% agarose gel electrophoresis to confirm the size and quality of the target DNA fragment. Subsequently, the target band was extracted from the electrophoresis gel using the AMPure XT bead recovery kit to achieve effective recovery of the PCR products.

[0031] 2.3 Library Construction and Sequencing PCR products were quality-assessed using an Agilent 2100 bioanalyzer and an Illumina library quantification kit; libraries with a concentration above 2 nM were considered acceptable. After confirming library quality, sequencing libraries from different samples (ensuring no repetition of index sequences) were serially diluted and mixed in appropriate proportions according to the expected sequencing volume. The mixed libraries were then denatured using NaOH to convert them into single-stranded DNA, ready for sequencing. Finally, a NovaSeq 6000 sequencer was used for 2 × 250 bp paired-end sequencing, employing the NovaSeq 6000 SP Reagent Kit (500 cycles).

[0032] 3. Bioinformatics analysis and biomarker screening, model construction 3.1 16S rDNA Sequencing Data Analysis Methods (1) Extraction of effective data First, the raw data of all samples were processed to extract valid data. Specific criteria included: removing primer sequences and balancing base sequences from the raw data; eliminating sequences shorter than 100 bp; excluding sequences containing more than 5% N (undefined ambiguous bases) after truncation; and removing chimeric sequences.

[0033] (2) Species annotation The DADA2 method in QIIME was used to perform length filtering and noise reduction on the raw data by running the denoise-paired command to obtain amplicon sequence variants (ASVs) and their abundance tables. During this process, ASVs that appeared only once were removed. Subsequently, the obtained ASV characteristic sequences were annotated using the SILVA database, and the NT-16S database was selected for comparison. Based on the ASV annotation results and abundance tables, species abundance tables at different taxonomic levels (including kingdom, phylum, class, order, family, genus, and species) were constructed.

[0034] (3) Species diversity analysis Alpha diversity refers to the diversity within a sample, typically quantified using indices such as Chao1, ACE, Simpson, Shannon, and Shannon's. Chao1 and ACE primarily reflect community richness, while Shannon and Simpson indices mainly reflect community diversity, and Shannon's index focuses on community evenness. Beta diversity reflects species differences between different samples, calculated based on Unifrac distance. High-dimensional distance matrices can be visualized using dimensionality reduction methods such as principal coordinates analysis (PCoA) or non-metric multidimensional scaling (NMDS). The Kruskal-Wallis test and PERMANOVA were used to compare the significance of inter-group differences in alpha and beta diversity, respectively.

[0035] (4) Differential species analysis and functional prediction This study screened differentially expressed oral microbiota using multiple methods: First, the Mann-Whitney U test was used to assess inter-group differences (Benjamini-Hochberg corrected p < 0.05); then, linear discriminant analysis (LDA) effect size (LEfSe) was used for multi-level discriminant analysis (Kruskal-Wallis screening followed by Wilcoxon test and LDA effect size assessment, threshold LDA > 3 and p < 0.05); simultaneously, a random forest model was constructed, and the construction method is as follows: Model construction and validation methods are as follows: 1) Data preprocessing and feature input: The relative abundance matrix at the genus / species level after screening is used as the input feature, and the sample group (case / control) is used as the binary outcome variable.

[0036] 2) Model parameter settings: Implemented using the randomForest package in R. The number of decision trees, ntree = 500, and the number of candidate variables for node splits, mtry, is set to the square root of the number of input variables by default. The remaining parameters use their default values ​​(nodesize = 1, replace = TRUE).

[0037] 3) Model Validation and Overfitting Control: 10-fold cross-validation was used to evaluate the model's generalization ability, and this was repeated 5 times to reduce bias caused by random partitioning. By plotting cross-validation error curves under different numbers of features, the optimal feature subset range was determined to ensure that the model did not overfit.

[0038] 4) Variable Importance Ranking: Calculate the average reduction in Gini coefficient for each microbial feature. A higher value indicates a greater contribution of the variable to the model's classification accuracy. The top 5 variables by importance are then output as key candidate biomarkers for subsequent joint diagnostic analysis.

[0039] The model's effectiveness was validated using Receiver Operating Characteristic (ROC) curves; finally, Venn diagram analysis was used to identify common species among the three methods as biomarkers.

[0040] In addition, PICRUSt2 (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2) was used to predict the functional profile of the oral microbiome from the Kyoto Encyclopedia of Genes and Genomes (KEGG), including KEGG Orthology and enzyme classification numbers, and the reliability of the prediction was assessed by NSTI value (threshold <0.15).

[0041] 3.2 Model Construction (1) Data preparation: The relative abundance values ​​of the core biomarker group in each sample are used as input features of the model.

[0042] (2) Logistic regression model: A joint diagnostic model was constructed using binary logistic regression. The PTDM disease status (yes / no) was used as the dependent variable, and the relative abundance of the core biomarkers was used as the independent variable. The regression equation was fitted by the maximum likelihood method to obtain the predicted probability value of each sample, i.e., the joint diagnostic index.

[0043] (3) Model validation: Use leave-one-out cross-validation or 10-fold cross-validation to evaluate the robustness of the model and prevent overfitting.

[0044] (4) Combined ROC curve plotting and AUC calculation.

[0045] (5) Obtaining the predicted probability: Substitute the abundance of the core biomarkers of each sample into the fitted Logistic regression model to calculate the predicted probability of each sample.

[0046] (6) ROC curve plotting: Using the predicted probability as the test variable and the actual prevalence of PTDM as the state variable, a receiver operating characteristic (ROC) curve was plotted. This curve reflects the sensitivity and specificity of the joint model at different diagnostic thresholds.

[0047] (7) AUC calculation: Calculate the area under the ROC curve (AUC) to evaluate the overall diagnostic efficacy of the combined model. The closer the AUC value is to 1, the stronger the discrimination ability of the combined model.

[0048] (8) Comparison of independent diagnostic efficacy To verify the advantages of the combined model over a single biomarker, ROC curves for each core biomarker's individual diagnosis were plotted and AUC was calculated, and compared with the AUC of the combined model.

[0049] 3.3 Results Analysis (1) Analysis of 16S rDNA V3-V4 region sequencing results 16S rDNA V3-V4 region sequencing was performed on 61 oral saliva samples, yielding a total of 4,610,037 high-quality sequences. The average sequencing depths in the NODAT group and the control group were 74,675 and 76,444 sequences, respectively, with no statistically significant difference between the groups (P>0.05). Figure 1 A).

[0050] Cluster analysis based on QIIME2 identified a total of 4909 ASVs. Venn diagrams showed that the PTDM group and the control group had 1591 and 1695 ASVs, respectively, for a total of 1623 ASVs. Figure 1 B).

[0051] (2) Analysis of microbial α and β diversity The results (Table 1) showed no statistically significant differences between the two groups in Shannon index (P=0.082), Simpson index (P=0.082), Chao1 index (P=0.650), and ace index (P=0.670). This indicates that there were no significant differences in the richness, diversity, and evenness of oral microbiota between the PTDM group and the control group. Furthermore, we analyzed the β-diversity of oral microbiota in the PTDM group and the control group. The results of PCoA and NMDS based on the weighted UniFrac distance matrix (see Table 1) show... Figure 2 There was no significant difference in oral microbiota between the PTDM group and the control group (PCoA: PCoA1, 43.24%, PCoA2, 19.02%; NMDS: Stress, 0.1267).

[0052] (3) Oral microbial composition of recipients in the PTDM group and the control group After processing and analyzing the 16S rDNA V3-V4 region sequencing data from 61 oral saliva samples, species abundance results at the phylum, class, order, family, and genus levels were obtained. A total of 827 phyla, 1503 classes, 3043 orders, 4797 families, and 8626 genera were annotated from the 61 samples. Among them, 30 samples from the PTDM group were annotated with 412 phyla, 747 classes, 1501 orders, 2408 families, and 4351 genera, while 31 samples from the control group were annotated with 415 phyla, 756 classes, 1542 orders, 2389 families, and 4275 genera.

[0053] To evaluate whether the differentially expressed oral microbiota at the genus level between the PTDM group and the control group could serve as a biomarker to distinguish PTDM recipients from normal glucose tolerance individuals, this study selected four differentially expressed oral microbiota from the intersection of the screened differentially expressed species. ROC curve analysis was then performed on these four differentially expressed microbiota to assess their discriminative performance.

[0054] The results (Table 2) showed that Vibrio succinate (AUC=0.810, P<0.001), Akkermania succinate (AUC=0.770, P=0.0138), Vibrio succinate (AUC=0.726, P<0.001), and Schwarzia succinate (AUC=0.733, P<0.001) were significantly different.

[0055] Next, we used a binary logistic regression model to calculate the total AUC value of these differentially expressed oral microbial biomarkers. The binary logistic regression is shown in Table 3, and the joint diagnostic model equation is: Y = -1.504 + 0.006 × Akkermansia relative abundance + 0.681 × Succinivibrio relative abundance - 0.521 × Succiniclasticum relative abundance + 9.072 × Schwartzia relative abundance.

[0056] The results are shown in Table 2. Figure 3 The overall oral microbiome model (AUC=0.981) was superior to single oral microbiome models in both the PTDM group and the control group; the sensitivity of the model was ≥100%, and the specificity was ≥96.6%.

[0057] (4) Construction of joint diagnostic model Using PTDM prevalence status (yes / no) as the dependent variable and the relative abundance of the core biomarker combination as the independent variable, a binary logistic regression analysis was employed. The joint diagnostic model equation was obtained by fitting the model using the maximum likelihood method: Y = -1.504 + 0.006 × Akkermansia spp. relative abundance + 0.681 × Vibrio succinate relative abundance - 0.521 × Vibrio succinate relative abundance + 9.072 × Schwarzia spp. relative abundance. The specific steps are as follows: 1) Collect oral saliva samples from kidney transplant recipients and divide them into case group and control group according to whether new diabetes develops after transplantation. Individuals with preoperative diabetes and stress-induced hyperglycemia are excluded. Before oral saliva sample collection, the recipients fasted for 1 hour, refrained from drinking water, alcohol, smoking, and chewing gum. Saliva samples were collected from the bilateral gingival margins and buccal mucosa and stored at -80℃ for 2 hours.

[0058] 2) Extraction of microbial DNA from saliva and PCR amplification and sequencing of the 16S rDNA V3-V4 region to obtain sequencing data; the primer sequences for PCR amplification were 341F: 5'-CCTACGGGNGGCWGCAG-3', 805R: 5'-GACTACHVGGGTATCTAATCC-3'; the PCR amplification program was: denaturation at 98℃ for 1 min, followed by 35 cycles of denaturation at 98℃ for 10 s, annealing at 54℃ for 30 s, extension at 72℃ for 45 s, and extension at 72℃ for 10 min.

[0059] 3) Data quality control, noise reduction and ASV clustering: species annotation was completed based on the database to obtain the microbial abundance matrix; DADA2 noise reduction and SILVA database annotation were used to calculate α diversity and β diversity and perform intergroup comparisons.

[0060] 4) Differential species were screened using difference tests, LEfSe analysis, and random forest models, and the intersection was taken to obtain the core oral microbial markers; the core oral microbial markers were Vibrio succinate, Akkermania succinate, Vibrio schwarzia succinate, and Vibrio succinate.

[0061] 5) A binary logistic regression joint diagnostic model was constructed using the relative abundance of core oral microbial biomarkers as input features. The binary logistic regression joint diagnostic model was constructed using binary logistic regression. Specifically, the PTDM disease status (yes / no) was used as the dependent variable, and the relative abundance of the core biomarkers was used as the independent variable. The regression equation was fitted using the maximum likelihood method to obtain the predicted probability value for each sample, i.e., the joint diagnostic index. The equation of the binary logistic regression joint diagnostic model is: Y = -1.504 + 0.006 × Akkermansia spp. relative abundance + 0.681 × Vibrio succinate spp. relative abundance - 0.521 × Vibrio succinate relative abundance + 9.072 × Schwarzia spp. relative abundance.

[0062] 6) Plot the ROC curve and calculate the AUC to evaluate the model's predictive performance.

[0063] The diagnostic threshold is Y ≥ -0.99 for PTDM positive and Y < -0.99 for PTDM negative; the model has AUC ≥ 0.981, sensitivity ≥ 100%, and specificity ≥ 96.6%.

[0064] In addition, the relative abundance values ​​of the 61 kidney transplant recipients are shown in Table 4.

[0065] Table 4. Relative abundance values ​​in 61 kidney transplant recipients patient Akkermansia Succinivibrio Succiniclasticum Schwartzia state Combined 1 263 0 0 0 yes 0.52 2 5 12 13 9 yes 82.02 3 5 23 12 8 yes 80.96 4 3 21 10 0 yes 8.06 5 4 14 6 3 yes 32.59 6 8 6 4 2 yes 19.14 7 9 14 12 6 yes 56.71 8 27 4 0 0 yes 1.83 9 12 4 0 5 yes 47.1 10 34 13 4 0 yes 5.92 11 35 7 4 11 yes 101.63 12 46 0 0 0 yes -0.78 13 40 11 4 9 yes 86.24 14 14 3 0 6 yes 55.51 15 31 0 0 2 yes 17.28 16 56 9 0 5 yes 50.77 17 0 23 0 0 yes 14.61 18 105 0 0 0 yes -0.42 19 3 37 9 0 yes 19.47 20 2 29 10 2 yes 31.64 21 0 42 13 2 yes 38.92 22 61 0 0 0 yes -0.69 23 3 19 10 0 yes 6.69 24 13 11 8 0 yes 2.35 25 55 0 0 0 yes -0.72 26 43 0 0 0 yes -0.8 27 29 0 0 0 yes -0.88 28 42 0 0 0 yes -0.8 29 0 3 8 7 yes 60.33 30 156 0 0 0 yes -0.12 31 0 14 21 0 no -2.46 32 0 0 0 0 no -1.05 33 0 0 0 0 no -1.05 34 9 0 0 0 no -1 35 0 0 0 0 no -1.05 36 251 2 0 0 no 1.81 37 0 0 0 0 no -1.05 38 0 0 0 0 no -1.05 39 0 0 0 0 no -1.05 40 3 0 0 0 no -1.04 41 0 0 0 0 no -1.05 42 3 0 0 0 no -1.04 43 4 0 0 0 no -1.03 44 5 0 0 0 no -1.02 45 0 0 0 0 no -1.05 46 10 0 0 0 no -0.99 47 9 0 0 0 no -1 48 0 0 0 0 no -1.05 49 0 0 0 0 no -1.05 50 3 0 0 0 no -1.04 51 4 0 0 0 no -1.03 52 0 0 0 0 no -1.05 53 3 0 0 0 no -1.04 54 0 0 0 0 no -1.05 55 0 0 0 0 no -1.05 56 12 0 0 0 no -0.98 57 8 0 0 0 no -1.01 58 7 0 0 0 no -1.01 59 7 0 0 0 no -1.01 60 403 0 0 0 no 1.36 61 7 0 0 0 no -1.01 4. Summary (1) The present invention performs ROC curve analysis on the core biomarker combination for each species. The results show that: Vibrio succinate AUC=0.810 (p<0.001), Akkermania succinate AUC=0.770 (p=0.0138), Vibrio succinate AUC=0.726 (p<0.001), and Schwarzia succinate AUC=0.733 (p<0.001). Each single biomarker has good diagnostic potential for PTDM.

[0066] (2) The robustness of the model was evaluated by 10-fold cross-validation, ROC curve was plotted and AUC was calculated. The results showed that the AUC of the combined diagnostic model was 0.981 (p<0.001), with a 95% confidence interval of 0.952~1.000, which was significantly better than the diagnostic efficacy of each single biomarker.

[0067] (3) The present invention constructs a logistic regression model for eight auxiliary biomarkers with high diagnostic potential. The results show that the extended model AUC=0.9774 (p<0.001), which also has extremely high diagnostic efficacy.

[0068] (4) Diagnosis of PTDM: Oral saliva samples were collected from kidney transplant recipients. Microbiome analysis was performed according to the method described in "2. Sequencing of 16S rDNA V3-V4 Regions of Microorganisms in Saliva Samples". The relative abundance of the core biomarker combination was calculated according to the method described in "3. Bioinformatics Analysis and Biomarker Screening". The relative abundance was substituted into the joint diagnostic model equation to calculate the Y value, and the predicted probability was calculated using a logistic regression function. In this embodiment, the diagnostic threshold was determined to be -0.99 using ROC curves. When Y ≥ -0.99, the result was considered PTDM positive; when Y < -0.99, the result was considered PTDM negative.

[0069] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A combination of microbial biomarkers for diagnosing diabetes mellitus after kidney transplantation, characterized in that, It consists of four core oral microbial markers: Succinivibrio, Akkermansia, Succiniclasticum, and Schwartzia; the markers are relative abundance at the genus level, and the test sample is saliva from the mouth of kidney transplant recipients. A diagnostic kit, characterized in that it comprises a reagent for detecting the combination of microbial markers of claim 1, the reagent comprising: (1) Saliva microbial DNA extraction reagent; (2) Primers for 16S rDNA V3-V4 region amplification: 341F:5'-CCTACGGGNGGCWGCAG-3' 805R: 5'-GACTACHVGGGTATCTAATCC-3'; (3) PCR amplification reagents, purification reagents, and sequencing library construction reagents; (4) Negative control, positive control and abundance quantification reference.

2. The use of the microbial biomarker combination of claim 1 or the kit of claim 2 in the preparation of a non-invasive diagnostic product for new-onset diabetes mellitus after kidney transplantation.

3. The application according to claim 3, characterized in that, Applications of the diagnostic products: (1) Detection of the V3-V4 region sequence of 16S rDNA from salivary microorganisms; (2) Calculate the relative abundance at the level of the four core microbial genera; (3) Substitute into the combined diagnostic model and output the PTDM positive / negative judgment result.

4. A method for constructing a predictive model for diabetes after kidney transplantation based on the oral microbiome, characterized in that, Includes the following steps: (1) Sample grouping: Oral saliva was collected from subjects ≥3 months after kidney transplantation and divided into PTDM case group and normal glucose tolerance control group, excluding individuals with preoperative diabetes and stress hyperglycemia; (2) DNA extraction and PCR amplification: Microbial DNA was extracted from saliva and the 16S rDNA V3-V4 region was amplified using primers 341F / 805R; (3) Sequencing and noise reduction: NovaSeq 6000 paired-end sequencing, DADA2 noise reduction, ASV clustering, SILVA database annotation, and obtaining the microbial abundance matrix; (4) Biomarker screening: Mann-Whitney U difference test, LEfSe analysis and random forest model were used to screen differential species, and the intersection was taken to obtain core oral microbial biomarkers; (5) Model construction: A binary logistic regression joint diagnostic model was constructed based on the relative abundance of four core oral microbial biomarkers. (6) Performance evaluation: construct ROC curves and calculate AUC, determine the diagnostic threshold, verify sensitivity, specificity, and accuracy, and evaluate the predictive performance of the model; The diagnostic threshold is -0.99; Y ≥ -0.99 is considered positive for PTDM, and Y < -0.99 is considered negative for PTDM. The model has an AUC ≥ 0.981, sensitivity ≥ 100%, and specificity ≥ 96.6%.

5. The construction method according to claim 5, characterized in that, Before the sample collection described in step (1), the subject fasted for 1 hour from food, water, alcohol, tobacco and chewing gum. Saliva samples were collected from the bilateral gingival margins and buccal mucosa and stored at -80℃ for 2 hours.

6. The construction method according to claim 5, characterized in that, The primer sequences for the PCR amplification were 341F: 5'-CCTACGGGNGGCWGCAG-3' and 805R: 5'-GACTACHVGGGTATCTAATCC-3'. The PCR amplification program was as follows: denaturation at 98℃ for 1 min, followed by 35 cycles of denaturation at 98℃ for 10 s, annealing at 54℃ for 30 s, extension at 72℃ for 45 s, and a final extension at 72℃ for 10 min.

7. The construction method according to claim 5, characterized in that, The core oral microbial markers are Vibrio succinate, Akkermania succinate, Vibrio succinate hydrophila, and Schwarzia succinate.

8. The construction method according to claim 5, characterized in that, The equation for the binary logistic regression joint diagnostic model is: Y = -1.504 + 0.006 × relative abundance of Akkermania + 0.681 × relative abundance of Vibrio succinate - 0.521 × relative abundance of Vibrio succinate + 9.072 × relative abundance of Schwarzschild.

Images