Establishment method and application of death prediction model of new type 2 diabetes mellitus

Abstract

The invention discloses a method for establishing a death prediction model of new type 2 diabetes mellitus. The method comprises the following steps: S1, collecting serum metabolites as a sample; s2, analyzing the sample to obtain a serum metabolite sample data set; s3, pre-processing the data to obtain a final serum metabolite sample data set; and S4, inputting the final serum metabolite sample data set into a new type 2 diabetes mellitus death prediction model to train the death prediction model, and obtaining a final new type 2 diabetes mellitus death prediction model after training is completed. The invention further discloses application of the death prediction model of the new type 2 diabetes mellitus in prediction of the occurrence probability of death of the new type 2 diabetes mellitus. The risk prediction accuracy is higher, operation and use are flexible, and the patient experience feeling and acceptability are high.

Description

Technical Field

[0001] The present invention belongs to the technical field of type 2 diabetes prognosis prediction, and specifically relates to a method for establishing and applying a death prediction model for newly-onset type 2 diabetes based on serum metabolites. Background Art

[0002] Diabetes is a chronic metabolic disease characterized by high blood sugar, mainly including type 1 diabetes and type 2 diabetes. Type 2 diabetes is a chronic disease caused by insufficient or reduced efficiency of insulin use, which is common in adults and is also called adult-onset diabetes. This disease is caused by genetics and environmental factors such as unhealthy lifestyles, overnutrition, and lack of physical activity. The initial symptoms are often mild, and many people are not discovered until complications occur or during routine physical examinations. Type 2 diabetes accounts for more than 90% of diabetes cases, and its risk of death from cardiovascular disease is nearly twice that of healthy individuals. Due to the heterogeneity between individuals in type 2 diabetes patients and differences in living environment or habits, the use of traditional clinical risk factors including gender, age, smoking history, glycated hemoglobin levels, etc., cannot meet the needs of early and accurate assessment of the risk of death from all causes, cardiovascular disease or non-cardiovascular disease.

[0003] Metabolomics can accurately detect metabolites or small molecule chemicals with a molecular mass of less than 1,500 Daltons in biological samples through high-throughput mass spectrometry and other methods. According to their chemical properties, they are divided into one of the following eight mutually exclusive chemical categories: amino acids and amino acid derivatives, carbohydrates, cofactors and vitamins, energy metabolism, lipids, nucleotides, peptides and exogenous substances. The system reflects the complex interaction of exogenous exposure, endogenous genetic background and intrinsic biochemical processes, comprehensively evaluates the human body's exogenous contact and response, and quantifies metabolites and metabolic pathways associated with diseases. With the integration of multi-dimensional technologies, metabolomics has gradually become an effective means to explore the molecular mechanisms and metabolic pathways of etiology, screen and identify disease biomarkers, and build disease risk prediction models.

[0004] Therefore, a metabolite-based prediction model for all-cause mortality, cardiovascular disease or non-cardiovascular disease mortality in new-onset type 2 diabetes is urgently needed to improve the accuracy of prediction. Summary of the invention

[0005] The purpose of the present invention is to provide a method for establishing and applying a death prediction model for newly-onset type 2 diabetes, so as to solve the problem proposed in the background technology that the use of traditional clinical risk factors including gender, age, smoking history, glycated hemoglobin level, etc., cannot meet the needs of early and accurate assessment of the risk of death from all causes, cardiovascular disease or non-cardiovascular disease.

[0006] To achieve the above object, the present invention provides a method for establishing a mortality prediction model for newly diagnosed type 2 diabetes, comprising the following steps:

[0007] S1, collect serum metabolites as samples;

[0008] S2. Analyze the sample to obtain a serum metabolite sample data set, wherein the serum metabolite sample data set includes M serum metabolite sample data subsets, each serum metabolite sample data subset includes abundance data of each component in the serum metabolite, and the components of the serum metabolite include 3,5-dichloro-2,6-dihydroxybenzoic acid;

[0009] S3, data pre-processing, performing pre-processing on the serum metabolite sample data set to obtain a final serum metabolite sample data set;

[0010] S4. Inputting the final serum metabolite sample data set into the new-onset type 2 diabetes death prediction model to train the death prediction model. After the training is completed, the final new-onset type 2 diabetes death prediction model is obtained for predicting the probability of death from new-onset type 2 diabetes.

[0011] In a specific embodiment, the new-onset type 2 diabetes death prediction model includes a new-onset type 2 diabetes death prediction model for all-cause death, a new-onset type 2 diabetes death prediction model for cardiovascular disease death, and a new-onset type 2 diabetes death prediction model for non-cardiovascular disease death;

[0012] The final new-onset type 2 diabetes death prediction model includes the final new-onset type 2 diabetes death prediction model for all-cause mortality, the final new-onset type 2 diabetes death prediction model for cardiovascular disease mortality, and the final new-onset type 2 diabetes death prediction model for non-cardiovascular disease mortality.

[0013] In a specific embodiment, the components of serum metabolites input into the final serum metabolite sample data set for the new-onset type 2 diabetes mortality prediction model for all-cause mortality are: 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine;

[0014] The components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid;

[0015] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, and dicarboxylic acid (C8-DC).

[0016] In a specific embodiment, the components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality are: 5-methyluridine (nucleothymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA);

[0017] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine;

[0018] The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, and tryptophan.

[0019] In a specific embodiment, the components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality are: 5-methyluridine (nucleothymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid;

[0020] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutanoic acid, and dihomo-linolenic acid (20:3n3 or n6);

[0021] The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositol, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoylglycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoylglycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoylglycerophosphatidylcholine (16:0 / 18:3n3), and 3-formyl indole.

[0022] In a specific embodiment, the components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality are: 5-methyluridine (nucleothymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphate Ethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid, hydroxy-N6,N6,N6-trimethyllysine, 3-amino-2-piperidone, 2-hydroxyoctanoic acid, 5α-pregnane-3β,20β-diol monosulfate, cystathionine;

[0023] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutanoic acid, dihomo-linolenic acid (20:3n3 or n6), dopamine-3-O-sulfate, vanillyl lactic acid, 5α-pregnane-3β,20β-diol monosulfate, 3,4-methyleneheptanoic acid, hydroxy-N6,N6,N6-trimethyllysine;

[0024] The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoyl glycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoyl glycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoyl glycerophosphatidylcholine (16:0 / 18:3n3), 3-formyl indole, mannoic acid, octadecenedioic acid (C18:1-DC), 1-methylnicotinamide, 1-palmitoleoyl-2-linoleoyl glycerophosphatidylcholine (16:1 / 18:2), 1-stearoyl-2-linoleoyl glycerophosphatidylethanolamine (18:0 / 18:2).

[0025] In a specific embodiment, in step S3, pre-processing is performed on the serum metabolite sample dataset to obtain a final serum metabolite sample dataset, including:

[0026] Serum metabolite quantification was accomplished by calculating the area under the chromatographic peak curve to obtain the raw peak area for each serum metabolite;

[0027] After obtaining the raw peak area of ​​each serum metabolite, it was standardized for subsequent statistical and bioinformatics analysis: first, the raw peak area of ​​each metabolite was logarithmically transformed with a base of 2 to reduce the skewed distribution of the overall value and make the data close to a normal distribution, and then normalized using the median. Finally, for serum metabolites with less than 20% missing values, the minimum value of all samples was used to fill the missing values.

[0028] In a specific embodiment, collecting serum metabolites in step S1 includes:

[0029] Methanol was added to the collected serum metabolite samples in a certain ratio; after oscillation and mixing, the samples were centrifuged; after centrifugation, the supernatant was taken from each sample and placed in a sample plate, blown dry with nitrogen, and then the reconstitution solution was added for subsequent detection and analysis.

[0030] The present invention also provides the application of the death prediction model for newly-onset type 2 diabetes mellitus established as described above in predicting the probability of death from newly-onset type 2 diabetes mellitus.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. The present invention provides a method for establishing a model of all-cause mortality, cardiovascular disease or non-cardiovascular disease mortality in newly-onset type 2 diabetes based on serum metabolites. Serum metabolites are collected to obtain a sample data set, and the sample data set is used to predict the long-term health of newly-onset type 2 diabetes. The risk prediction accuracy is higher, serum metabolites are minimally invasively collected, the operation is flexible, and the patient experience and acceptance are high.

[0033] 2. The present invention selects 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), and ergothioneine from the serum metabolite components to further improve the accuracy of predicting the risk of all-cause mortality.

[0034] The selection of 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazoleacetic acid, indolepropionic acid, and 3,5-dichloro-2,6-dihydroxybenzoic acid among the serum metabolite components further improved the accuracy of predicting the risk of death from cardiovascular disease.

[0035] The selection of 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, and dicholic acid (C8-DC) among the serum metabolite components further improved the accuracy of predicting the risk of death from non-cardiovascular diseases.

[0036] 3. In the present invention, the sample data of newly-emerged type 2 diabetes metabolites are input, and the data can be flexibly selected according to the prediction accuracy requirements. When the accuracy requirements are high, a combination of more components is selected as the sample data, and when the accuracy requirements are low, a combination of fewer components is selected as the sample data, thereby improving the convenience and ease of use of the prediction model.

[0037] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention is further described in detail below. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0039] Figure 1 It is a flow chart of a method for establishing a death prediction model for newly-onset type 2 diabetes of the present invention;

[0040] Figure 2 is a volcano plot of metabolites highly correlated with all-cause mortality identified in a training set in an embodiment of the present invention;

[0041] Figure 3is a volcano plot of metabolites highly correlated with cardiovascular disease mortality identified in a training set in an embodiment of the present invention;

[0042] Figure 4 is a volcano plot of metabolites highly correlated with non-cardiovascular disease mortality identified in a training set in an embodiment of the present invention;

[0043] Figure 5 This is a graph showing the AUCs value of all-cause mortality of newly diagnosed type 2 diabetes obtained using a combination of four serum metabolite components in the training set and eight traditional risk factors in an embodiment of the present invention;

[0044] Figure 6 This is a graph showing the AUCs value of the all-cause mortality of newly diagnosed type 2 diabetes obtained using a combination of four serum metabolite components in an external validation set and eight traditional risk factors in an embodiment of the present invention;

[0045] Figure 7 This is a graph showing the AUCs values ​​of the cardiovascular disease mortality results of newly diagnosed type 2 diabetes mellitus obtained using a combination of four serum metabolite components in the training set and eight traditional risk factors in an embodiment of the present invention;

[0046] Figure 8 This is a graph showing the AUCs values ​​of the cardiovascular disease mortality results of newly diagnosed type 2 diabetes mellitus obtained using a combination of four serum metabolite components in an external validation set and eight traditional risk factors in an embodiment of the present invention;

[0047] Fig. 9 This is a graph showing the AUCs values ​​of the new-onset type 2 diabetes non-cardiovascular disease mortality results obtained using a combination of four serum metabolite components in the training set and eight traditional risk factors in an embodiment of the present invention;

[0048] Fig.10 This is a graph showing the AUCs values ​​of the outcomes of non-cardiovascular disease mortality in newly diagnosed type 2 diabetes obtained using a combination of four serum metabolite components in an external validation set and eight traditional risk factors in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The embodiments of the present invention are described in detail below. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] The present invention provides a method for establishing a mortality prediction model for newly diagnosed type 2 diabetes, the method comprising the following steps:

[0051] S1. Collect serum metabolites as samples.

[0052] Participant cohort: Based on the new-onset type 2 diabetes patients in the national multicenter new-onset diabetes cohort (covering 46 hospital centers in seven geographical regions of China), the death status, death date and cause of death of the study subjects were determined by linking the Chinese Center for Disease Control and Prevention (CDC) and the National Cancer Center. By excluding those who died due to accidents and those without serum samples, the new-onset type 2 diabetes patients aged 20-80 years who met the above conditions and had a death event were screened as the case group, totaling 152 participants. 152 new-onset type 2 diabetes patients who did not have a death event were matched according to the recruitment center, patient gender, age, and serum sample collection time in a 1:1 condition. A total of 304 study subjects formed a nested case-control pair. Specifically, the training cohort included 212 participants from 26 recruitment centers, and the validation cohort included 92 participants from 8 recruitment centers. The model was trained on the training set, and the model constructed by screening was tested on the external validation set.

[0053] When the cohort was recruited, general information and clinical data of all participants were collected through questionnaires, including age, gender, body mass index, and smoking history. At the same time, some peripheral marker information, including glycated hemoglobin level and triglyceride level, was collected by collecting fasting blood samples from participants.

[0054] For the national multicenter new-onset diabetes cohort, inclusion criteria included: (1) meeting the World Health Organization (WHO) 1999 diabetes diagnostic criteria; (2) age of onset ≥18 years; (3) diabetes duration <1 year. Exclusion criteria included: (1) gestational or gestational diabetes; (2) acute diseases that may affect glucose metabolism; (3) common malignant tumors; (4) diabetes secondary to other causes; (5) diabetes with unclear classification.

[0055] Collection of serum metabolites: Serum metabolites were collected before breakfast (usually before 8 a.m.) on the day of recruitment into the National Multicenter New-onset Diabetes Cohort, after fasting for 8 to 10 hours, to exclude the effects of short-term diet on serum metabolites. Fasting venous blood was collected into blood collection tubes containing inert separation gel and coagulant. The blood samples were allowed to stand at room temperature for stratification, and then centrifuged at 3500 rpm for 15 minutes at 4°C. The supernatant was transferred to a new centrifuge tube and frozen at -80°C. Before testing, the samples were thawed at 4°C.

[0056] The collected serum metabolites are pre-processed, and the pre-processing includes:

[0057] Methanol was added to the collected serum metabolite samples at a ratio of 1:4. After oscillation for 3 minutes to mix, insoluble matter was precipitated by centrifugation at 4000×g for 10 minutes at 20°C. After centrifugation, 4 portions of 100 μL of supernatant were taken from each sample and placed in 4 sample plates, blown dry with nitrogen, and then added with a reconstitution solution for subsequent UPLC-MS / MS detection and analysis. The reconstitution solution was water, acetonitrile and / or methanol.

[0058] The method for establishing a death prediction model for newly-onset type 2 diabetes provided by the present invention collects serum metabolites, obtains a sample data set, and uses the sample data set to predict the probability of death from newly-onset type 2 diabetes. It is non-destructive to the human body and, compared with other destructive tests, improves patient experience and acceptance.

[0059] S2. Analyze the sample to obtain a serum metabolite sample data set, wherein the serum metabolite sample data set includes M serum metabolite sample data subsets, each serum metabolite sample data subset contains abundance data of each component in the serum metabolites, the components of the serum metabolites cover 736 known metabolites, and the components of the serum metabolites include 3,5-dichloro-2,6-dihydroxybenzoic acid.

[0060] Chromatographic and mass spectrometric methods: Data collection used four different UPLC-MS / MS methods, all of which were carried out using ACQUITY2DUPLC (Waters, Milford, MA, USA) ultra-high performance liquid chromatography combined with Q Exactive (QE) high-resolution mass spectrometry (Thermo Fisher Scientific, San Jose, USA). The QE mass spectrometry parameters were a scan resolution of 35,000 and a scan range of 70-1000 m / z. The specific parameters of the four UPLC-MS / MS methods are as follows:

[0061] Method 1: QE was detected in positive ion electrospray ionization (ESI) mode, and the liquid phase was separated using a C18 column (UPLCBEH C18, 2.1x100mm, 1.7um; Waters), with the mobile phases being water (A) and methanol (B) containing 0.05% PFPA and 0.1% FA;

[0062] Method 2: QE was detected in negative ion electrospray ionization (ESI) mode, and the liquid phase was separated using a C18 column (UPLCBEH C18, 2.1x100mm, 1.7um; Waters), with the mobile phases being water (A) and methanol (B) containing 6.5mM ammonium bicarbonate;

[0063] Method 3: QE was detected in positive ion electrospray ionization (ESI) mode, and the liquid phase was separated using a C18 column (UPLCBEH C18, 2.1x100mm, 1.7um; Waters), with the mobile phases being water (A) and methanol / acetonitrile / water (B) containing 0.05% PFPA and 0.1% FA;

[0064] Method 4: QE was detected in negative ion electrospray ionization (ESI) mode, and the liquid phase was separated using a HILIC column (UPLC BEH Amide, 2.1x150mm, 1.7um; Waters), with the mobile phase consisting of water (A) and acetonitrile (B) containing 10mM ammonium formate.

[0065] The new-onset type 2 diabetes death prediction model includes the new-onset type 2 diabetes death prediction model for all-cause mortality, the new-onset type 2 diabetes death prediction model for cardiovascular disease mortality, and the new-onset type 2 diabetes death prediction model for non-cardiovascular disease mortality.

[0066] The components of serum metabolites input into the final serum metabolite sample data set of the new-onset type 2 diabetes mortality prediction model for all-cause mortality of the present invention are: 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine;

[0067] The components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid;

[0068] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, and dicarboxylic acid (C8-DC).

[0069] Preferably, the components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality of the present invention are: 5-methyluridine (nuclear thymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA);

[0070] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine;

[0071] The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, and tryptophan.

[0072] More preferably, the components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality of the present invention are: 5-methyluridine (nuclear thymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid;

[0073] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutanoic acid, and dihomo-linolenic acid (20:3n3 or n6);

[0074] The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositol, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoylglycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoylglycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoylglycerophosphatidylcholine (16:0 / 18:3n3), and 3-formyl indole.

[0075] Further preferably, the components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes death prediction model for all-cause death of the present invention are: 5-methyluridine (nuclear thymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine ethyl phosphate Alkylamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid, hydroxy-N6,N6,N6-trimethyllysine, 3-amino-2-piperidone, 2-hydroxyoctanoic acid, 5α-pregnane-3β,20β-diol monosulfate, cystathionine;

[0076] The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutanoic acid, dihomo-linolenic acid (20:3n3 or n6), dopamine-3-O-sulfate, vanillyl lactic acid, 5α-pregnane-3β,20β-diol monosulfate, 3,4-methyleneheptanoic acid, hydroxy-N6,N6,N6-trimethyllysine;

[0077] The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoyl glycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoyl glycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoyl glycerophosphatidylcholine (16:0 / 18:3n3), 3-formyl indole, mannoic acid, octadecenedioic acid (C18:1-DC), 1-methylnicotinamide, 1-palmitoleoyl-2-linoleoyl glycerophosphatidylcholine (16:1 / 18:2), 1-stearoyl-2-linoleoyl glycerophosphatidylethanolamine (18:0 / 18:2).

[0078] In the present invention, sample data of the prediction model for all-cause mortality, cardiovascular disease or non-cardiovascular disease mortality of newly-onset type 2 diabetes can be input, and can also be flexibly selected according to the prediction accuracy requirements. When the accuracy requirement is high, a combination of more components is selected as sample data, and when the accuracy requirement is low, a combination of fewer components is selected as sample data, thereby improving the convenience and ease of use of the prediction model.

[0079] S3. Data pre-processing: performing pre-processing on the serum metabolite sample data set to obtain a final serum metabolite sample data set.

[0080] Specifically, the test data is pre-processed, and after confirming that it meets the quality control requirements, the chromatographic peak is extracted using the Hangzhou Kailaipu data processing system. Then, by comparing the metabolite database built based on the standard, the retention index, the primary accurate mass number and the secondary fragment spectrum are used to extract the peak. Figure 3 Metabolite identification was performed by strictly matching the indicators. Serum metabolite quantification was completed by calculating the area under the chromatographic peak curve to obtain the original peak area of ​​each serum metabolite.

[0081] After obtaining the raw peak area of ​​each serum metabolite, the raw peak area of ​​each metabolite was logarithmically transformed with a base of 2 to reduce the skewed distribution of the overall value and make the data close to a normal distribution, and then normalized using the median. Finally, for serum metabolites with less than 20% missing values, the minimum value of all samples was used to fill the missing values.

[0082] S4. Inputting the final serum metabolite sample data set into the new-onset type 2 diabetes death prediction model to train the death prediction model. After the training is completed, the final new-onset type 2 diabetes death prediction model is obtained for predicting the probability of death from new-onset type 2 diabetes.

[0083] The final new-onset type 2 diabetes death prediction model includes the final new-onset type 2 diabetes death prediction model for all-cause mortality, the final new-onset type 2 diabetes death prediction model for cardiovascular disease mortality, and the final new-onset type 2 diabetes death prediction model for non-cardiovascular disease mortality.

[0084] Example 1

[0085] In this embodiment, a total of 304 research subjects from 34 centers in the national multicenter new-onset diabetes cohort were included in the nested case-control pairs. The death status, date of death and cause of death of the research subjects were determined by linking the Chinese Center for Disease Control and Prevention (CDC) and the National Cancer Center. Specifically, 152 people with deaths and 152 matched controls without deaths were included; among them, the training cohort included 212 participants from 26 recruitment centers, and the validation cohort included 92 participants from 8 recruitment centers. As shown in Table 1, in the training set and the external validation set data set, there was no significant difference in age, gender, smoking history, educational level, body mass index, glycated hemoglobin level, etc. between patients with type 2 diabetes who had deaths and patients with type 2 diabetes who had not had deaths.

[0086] Table 1. Clinical information of case and control participants in the training set and internal validation set:

[0087]

[0088] Serum metabolite profile:

[0089] Four different UPLC-MS / MS methods were used for data acquisition, all of which were carried out using ACQUITY 2D UPLC (Waters, Milford, MA, USA) ultra-high performance liquid chromatography combined with Q Exactive (QE) high-resolution mass spectrometry (Thermo Fisher Scientific, San Jose, USA). The QE mass spectrometry parameters were a scan resolution of 35,000 and a scan range of 70-1000 m / z. After preprocessing of the mass spectrometry data, the metabolite database constructed based on the standard was compared and the retention index, primary accurate mass number and secondary fragmentation spectrum were used to identify the metabolites. Figure 3 Metabolite identification was performed by strictly matching the indicators. Metabolite quantification was completed by calculating the area under the chromatographic peak curve to obtain the raw peak area of ​​each metabolite. The raw peak area of ​​each metabolite was logarithmically transformed with a base of 2 to reduce the skewed distribution of the overall values ​​and make the data close to normal distribution, and then normalized using the median, and finally the minimum value of all samples was used to fill the missing values. A total of 736 metabolites were detected, which were divided into one of the following mutually exclusive chemical categories according to their chemical characteristics: amino acids and amino acid derivatives, carbohydrates, cofactors and vitamins, energy metabolism, lipids, nucleotides, peptides, and exogenous substances. The median and interquartile range of the coefficient of variation and intraclass correlation coefficient of each metabolite were 0.85% (0.51%-1.42%) and 0.84 (0.83-0.85), respectively. The metabolite detection results were highly consistent, and the experimental data were highly reliable.

[0090] Based on the training set, multivariate conditional logistic regression was used to evaluate the strength of association between the relative levels of 736 known metabolites and cardiovascular disease mortality when adjusting for covariates including recruitment center, gender, age, body mass index, glycated hemoglobin, systolic blood pressure, diastolic blood pressure, triglycerides, and smoking status. Based on the above evaluation results, P < 0.05 was set as highly correlated, and 119 metabolites highly correlated with all-cause mortality were obtained ( Figure 2 ); 57 metabolites highly associated with cardiovascular disease mortality ( Figure 3 ); and 59 metabolites highly associated with non-cardiovascular disease mortality ( Figure 4 ).

[0091] Establishment of a model for all-cause mortality, cardiovascular disease or non-cardiovascular disease mortality in new-onset type 2 diabetes based on serum metabolites:

[0092] For all 119 metabolites identified that were highly associated with all-cause mortality ( Figure 2 ); 57 metabolites highly associated with cardiovascular disease mortality ( Figure 3) and 59 metabolites highly associated with non-cardiovascular disease mortality ( Figure 4 ) performed machine learning SHAP value calculations based on XGBOOST to determine the contribution of the prediction model based on serum metabolites. Metabolites with higher contributions were screened, and a prediction model was constructed in combination with eight traditional risk factors including age, gender, body mass index, glycated hemoglobin, systolic blood pressure, diastolic blood pressure, triglycerides, and smoking status. The model prediction ability was evaluated by calculating the area under the curve AUC and its 95% credible interval. When the area under the curve was 0.5-0.69, it indicated that the model discrimination ability was poor, 0.7-0.79 indicated that the model discrimination ability was acceptable, and >0.8 indicated that the model discrimination ability was excellent. The generalization ability of the confirmed model was formed.

[0093] For all-cause mortality:

[0094] Specifically, four combinations of serum metabolite components were set up, that is, the combinations of 5, 10, 15, and 20 serum metabolite components with the highest SHAP absolute values ​​were selected to distinguish between case and control subjects. Figure 5 shown.

[0095] In the training set, the area under the curve value and confidence interval of the traditional risk factor (8) model were 0.65 (0.57, 0.72), and the model had poor discrimination ability. When the combination of the four serum metabolite components was combined with the traditional risk factors (8), the area under the curve (AUCs) value was 0.75-0.82. The more serum metabolite components were included, the larger the AUCs area. The confidence interval of the combination of 5 serum metabolite components was 0.69-0.82, and the confidence interval of the combination of 20 serum metabolite components was 0.77-0.88, indicating that the model had better discrimination ability and significantly higher identification ability than the traditional risk factor model (P values ​​<0.005).

[0096] The external validation set results are as follows Figure 6 As shown in the figure, the area under the curve value and confidence interval of the traditional risk factor (8) model are 0.69 (0.58, 0.72), and the model has poor discrimination ability. When the combination of the four serum metabolite components is combined with the traditional risk factors (8), the area under the curve (AUCs) value is 0.81-0.95. The AUCs values ​​of the training set and the external validation set shown in the above results are all above 0.75, which shows that the prediction tool based on the serum metabolome performs well in patients with type 2 diabetes, and the generalization ability of the model has been confirmed.

[0097] A combination of 5 serum metabolite components, wherein the serum metabolite components are 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), and ergothioneine.

[0098] A combination of 10 serum metabolite components, including the following serum metabolite components: 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoylsphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, and dimethylarginine (SDMA+ADMA).

[0099] A combination of 15 serum metabolite components, including the following serum metabolite components: 5-methyluridine (nuclear thymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, and 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid.

[0100] The composition of 20 serum metabolites includes: 5-methyluridine (nucleothymidine), 3-sulfatide chenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoyl glycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoyl sphingosine phosphoethanolamine (d18:1 / 16:0), sulfur Acid cholesterol, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid, hydroxy-N6,N6,N6-trimethyllysine, 3-amino-2-piperidone, 2-hydroxyoctanoic acid, 5α-pregnane-3β,20β-diol monosulfate, cystathionine.

[0101] For cardiovascular death:

[0102] Specifically, four combinations of serum metabolite components were set up, that is, the combinations of 5, 10, 15, and 20 serum metabolite components with the highest SHAP absolute values ​​were selected to distinguish between case and control subjects. Figure 7 shown.

[0103] In the training set, the area under the curve value and confidence interval of the traditional risk factor (8) model were 0.67 (0.57, 0.78), and the model had poor discrimination ability. When the combination of the four serum metabolite components was combined with the traditional risk factors (8), the area under the curve (AUCs) value was 0.84-0.91. The more serum metabolite components were included, the larger the AUCs area. The confidence interval of the combination of 5 serum metabolite components was 0.76-0.91, and the confidence interval of the combination of 20 serum metabolite components was 0.86-0.97, indicating that the model had better discrimination ability and significantly higher identification ability than the traditional risk factor model (P values ​​<0.005).

[0104] The external validation set results are as follows Figure 8 As shown in the figure, the area under the curve value and confidence interval of the traditional risk factor (8) model are 0.64 (0.49, 0.80), and the model has poor discrimination ability. When the combination of the four serum metabolite components is combined with the traditional risk factors (8), the area under the curve (AUCs) value is 0.79-0.99. The AUCs values ​​of the training set and the external validation set shown in the above results are all above 0.85, which shows that the prediction tool based on the serum metabolome performs well in patients with type 2 diabetes, and the generalization ability of the model has been confirmed.

[0105] A combination of 5 serum metabolite components, wherein the serum metabolite components are 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, and 3,5-dichloro-2,6-dihydroxybenzoic acid.

[0106] A combination of 10 serum metabolite components, including the following serum metabolite components: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, and 3-methoxytyrosine.

[0107] A combination of 15 serum metabolite components, including the following serum metabolite components: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazoleacetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutyric acid, and dihomo-linolenic acid (20:3n3 or n6).

[0108] A combination of 20 serum metabolite components, including the following serum metabolite components: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutyric acid, dihomo-linolenic acid (20:3n3 or n6), dopamine-3-O-sulfate, vanillyl lactic acid, 5α-pregnane-3β,20β-diol monosulfate, 3,4-methyleneheptanoic acid, hydroxy-N6,N6,N6-trimethyllysine.

[0109] For non-cardiovascular death:

[0110] Specifically, four combinations of serum metabolite components were set up, that is, the combinations of 5, 10, 15, and 20 serum metabolite components with the highest SHAP absolute values ​​were selected to distinguish between case and control subjects. Fig. 9 shown.

[0111] In the training set, the area under the curve value and confidence interval of the traditional risk factor (8) model were 0.66 (0.56, 0.76), and the model had poor discrimination ability. When the combination of the four serum metabolite components was combined with the traditional risk factors (8), the area under the curve (AUCs) value was 0.80-0.90. The more serum metabolite components were included, the larger the AUCs area. The confidence interval of the combination of 5 serum metabolite components was 0.72-0.88, and the confidence interval of the combination of 20 serum metabolite components was 0.84-0.95, indicating that the model had better discrimination ability and significantly higher identification ability than the traditional risk factor model (P values ​​<0.05).

[0112] The external validation set results are as follows Fig.10As shown in the figure, the area under the curve value and confidence interval of the traditional risk factor (8) model are 0.73 (0.57, 0.88), and the model discrimination ability is acceptable. When the combination of the four serum metabolite components is combined with the traditional risk factors (8), the area under the curve (AUCs) value can reach 0.85-0.99. The AUCs values ​​of the training set and the external validation set shown in the above results are all above 0.80, which shows that the prediction tool based on the serum metabolome performs well in patients with type 2 diabetes, and the generalization ability of the model has been confirmed.

[0113] A combination of 5 serum metabolite components, wherein the serum metabolite components are 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC)

[0114] A combination of 10 serum metabolite components, wherein the serum metabolite components are: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, and tryptophan.

[0115] A combination of 15 serum metabolite components, including the following serum metabolite components: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositol, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoylglycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoylglycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoylglycerophosphatidylcholine (16:0 / 18:3n3), and 3-formyl indole.

[0116] The combination of 20 serum metabolite components, including the serum metabolite components: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositol, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoylglycerophosphatidylethanolamine (16:0 / 18:1), 1 -oleoyl-2-docosahexaenoyl glycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoyl glycerophosphatidylcholine (16:0 / 18:3n3), 3-formyl indole, mannoic acid, octadecenedioic acid (C18:1-DC), 1-methylnicotinamide, 1-palmitoleoyl-2-linoleoyl glycerophosphatidylcholine (16:1 / 18:2), 1-stearoyl-2-linoleoyl glycerophosphatidylethanolamine (18:0 / 18:2).

[0117] Example 2

[0118] The present invention also provides an application of a death prediction model for newly-onset type 2 diabetes established by any one of the above-mentioned establishment methods in predicting the probability of death from newly-onset type 2 diabetes.

[0119] In specific applications, the new-onset type 2 diabetes death prediction model for all-cause death, the new-onset type 2 diabetes death prediction model for cardiovascular disease death, and the new-onset type 2 diabetes death prediction model for non-cardiovascular disease death are used for prediction. First, the patient's serum metabolites are collected and analyzed to obtain the abundance data of the corresponding components, which are then imported into the corresponding three final models for prediction to obtain the prediction results.

[0120] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for establishing a mortality prediction model for newly diagnosed type 2 diabetes, characterized in that: The steps include: S1, collect serum metabolites as samples; S2. Analyze the sample to obtain a serum metabolite sample data set, wherein the serum metabolite sample data set includes M serum metabolite sample data subsets, each serum metabolite sample data subset includes abundance data of each component in the serum metabolite, and the components of the serum metabolite include 3,5-dichloro-2,6-dihydroxybenzoic acid; S3, data pre-processing, performing pre-processing on the serum metabolite sample data set to obtain a final serum metabolite sample data set; S4. Inputting the final serum metabolite sample data set into the new-onset type 2 diabetes death prediction model to train the death prediction model. After the training is completed, the final new-onset type 2 diabetes death prediction model is obtained for predicting the probability of death from new-onset type 2 diabetes.

2. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 1, characterized in that: The new-onset type 2 diabetes death prediction model includes a new-onset type 2 diabetes death prediction model for all-cause death, a new-onset type 2 diabetes death prediction model for cardiovascular disease death, and a new-onset type 2 diabetes death prediction model for non-cardiovascular disease death; The final new-onset type 2 diabetes death prediction model includes the final new-onset type 2 diabetes death prediction model for all-cause mortality, the final new-onset type 2 diabetes death prediction model for cardiovascular disease mortality, and the final new-onset type 2 diabetes death prediction model for non-cardiovascular disease mortality.

3. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 2, characterized in that: The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for all-cause mortality were: 5-methyluridine (nucleothymidine), 3-sulfatochenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-enyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine; The components of serum metabolites in the final serum metabolite sample data set input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid; The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, and dicarboxylic acid (C8-DC).

4. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 2, characterized in that: The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality were: 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoylsphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA); The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine; The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, and tryptophan.

5. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 2, characterized in that: The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for all-cause mortality were: 5-methyluridine (nucleothymidine), 3-sulfatidechenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoylsphingosine phosphoethanolamine (d18:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoylorotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid; The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutanoic acid, and dihomo-linolenic acid (20:3n3 or n6); The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositol, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoylglycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoylglycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoylglycerophosphatidylcholine (16:0 / 18:3n3), and 3-formyl indole.

6. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 2, characterized in that: The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for all-cause mortality were: 5-methyluridine (nucleothymidine), 3-sulfatochenodeoxycholic acid glycine, 3,5-dichloro-2,6-dihydroxybenzoic acid, 1-(1-alkenyl-palmitoyl)-2-oleoylglycerophosphatidylcholine (P-16:0 / 18:1), ergothioneine, lactose, palmitoylsphingosine phosphoethanolamine (d1 8:1 / 16:0), cholesterol sulfate, 1,5-anhydroglucitol, dimethylarginine (SDMA+ADMA), epiandrosterone sulfate, perfluorooctane sulfonate, carbamoyl orotic acid, phenylacetylglutamine, 3-carboxy-4-methyl-5-propyl-2-furanpropionic acid, hydroxy-N6,N6,N6-trimethyllysine, 3-amino-2-piperidone, 2-hydroxyoctanoic acid, 5α-pregnane-3β,20β-diol monosulfate, cystathionine; The components of serum metabolites in the final serum metabolite sample dataset input into the new-onset type 2 diabetes mortality prediction model for cardiovascular disease mortality were: 2-hydroxyphytanate, N-acetyl-1-methylhistidine, 1-ribosyl imidazole acetic acid, indolepropionic acid, 3,5-dichloro-2,6-dihydroxybenzoic acid, galactonic acid, erythritol / threotetraol, dodecadienoic acid (12:2), 5α-androstane-3α,17β-diol monosulfate, 3-methoxytyrosine, 2'-deoxyuridine, fructose, palmitoyl sphingomyelin (d18:1 / 16:0), 2,3-dihydroxy-2-methylbutanoic acid, dihomo-linolenic acid (20:3n3 or n6), dopamine-3-O-sulfate, vanillyl lactic acid, 5α-pregnane-3β,20β-diol monosulfate, 3,4-methyleneheptanoic acid, hydroxy-N6,N6,N6-trimethyllysine; The components of serum metabolites in the final serum metabolite sample dataset input into the prediction model for new-onset type 2 diabetes mortality for non-cardiovascular disease mortality were: 3,5-dichloro-2,6-dihydroxybenzoic acid, N-acetylproline, 6-bromotryptophan, 3β,7α-dihydroxy-5-cholestenic acid, diosminic acid (C8-DC), arachidic acid (20:0), stachydrine, glycerophosphoinositide, N-acetylhexosamine, tryptophan, docosyl sphingomyelin (d18:1 / 22:0), 1-palmitoyl-2-oleoyl glycerophosphatidylethanolamine (16:0 / 18:1), 1-oleoyl-2-docosahexaenoyl glycerophosphatidylcholine (18:1 / 22:6), 1-palmitoyl-2-α-linolenoyl glycerophosphatidylcholine (16:0 / 18:3n3), 3-formyl indole, mannoic acid, octadecenedioic acid (C18:1-DC), 1-methylnicotinamide, 1-palmitoleoyl-2-linoleoyl glycerophosphatidylcholine (16:1 / 18:2), 1-stearoyl-2-linoleoyl glycerophosphatidylethanolamine (18:0 / 18:2).

7. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 1, characterized in that: In step S3, pre-processing is performed on the serum metabolite sample data set to obtain a final serum metabolite sample data set, including: Serum metabolite quantification was accomplished by calculating the area under the chromatographic peak curve to obtain the raw peak area for each serum metabolite; After obtaining the raw peak area of ​​each serum metabolite, it was standardized for subsequent statistical and bioinformatics analysis: first, the raw peak area of ​​each metabolite was logarithmically transformed with a base of 2 to reduce the skewed distribution of the overall value and make the data close to a normal distribution, and then normalized using the median. Finally, for serum metabolites with less than 20% missing values, the minimum value of all samples was used to fill the missing values.

8. The method for establishing a mortality prediction model for newly diagnosed type 2 diabetes according to claim 1, characterized in that: The step S1 of collecting serum metabolites comprises: Methanol was added to the collected serum metabolite samples in a certain ratio; after oscillation and mixing, the samples were centrifuged; after centrifugation, the supernatant was taken from each sample and placed in a sample plate, blown dry with nitrogen, and then the reconstitution solution was added for subsequent detection and analysis.

9. Use of the death prediction model for newly-onset type 2 diabetes established as claimed in any one of claims 1 to 8 in predicting the probability of death from newly-onset type 2 diabetes.

Images