Use of follistatin in predicting risk of type 2 diabetes

Abstract

The present disclosure details and reports herein the use of follistatin, GCKR-regulated hepatic follistatin secretion, as a biomarker for the early diagnosis and / or prediction of type 2 diabetes. Additionally, disclosed herein are methods for constructing a biomarker signature for the early prediction of type 2 diabetes in humans.

Description

[Technical field]

[0001] Technical Field HbA 1c Presented herein is a novel means for type 2 diabetes risk assessment that incorporates three known blood biomarkers for type 2 diabetes including insulin, proinsulin, and C-peptide, using follistatin as a marker in a model with good predictive value up to four years prior to disease onset. [Background technology]

[0002] background The absolute global economic burden of diabetes is estimated to increase from US$ 1.3 trillion in 2015 to US$ 2.5 (2.4–2.6) trillion by 2030, which is equivalent to 1.2% of global GDP. [1] In the United States alone, the average medical expenditure for a patient with diagnosed diabetes is $16,752, about 2.3 times higher than the expenditure for those without diabetes. In some health systems, diabetes care accounts for as much as 25% of overall costs.

[0003] It is possible to prevent type 2 diabetes (T2D) through lifestyle interventions if disease risk can be detected early enough. [2]Currently, oral glucose tolerance tests (OGTT) and fasting plasma glucose (FPG) are used to assess diabetes risk. However, diabetes and even complications may already have occurred when abnormal glucose levels are detected. Furthermore, as diabetes is a systemic disease, multiple blood signatures may be altered as a result, which makes the assessment of diabetes risk based on glucose alone questionable. Nevertheless, in current clinical practice, the diagnosis of potential diabetes is assessed only by glucose measurements: patients either have high blood glucose levels (diabetic) or normal blood glucose levels (non-diabetic). However, even among "non-diabetic", each individual may have a different risk level of future diabetes development, which cannot be efficiently assessed by currently available techniques and biological marker measurements. Therefore, it has proven appropriate to assess type 2 diabetes risk by a multivariate individualized risk score. Summary of the Invention [Means for solving the problem]

[0004] In the present report and study, the inventors report the development of a tool to cluster diabetes-free individuals into different risk groups that incorporates different biomarkers. Furthermore, the inventors have established a mathematical model that can accurately predict diabetes risk.

[0005] Surprisingly, it has been found that follistatin can be used as a biomarker for the early diagnosis of type 2 diabetes, and its use is reported herein.Furthermore, a method for constructing a biomarker signature for the early prediction of type 2 diabetes in humans is disclosed herein. [Brief description of the drawings]

[0006] BRIEF DESCRIPTION OF THE DRAWINGS [Figure 1] Hepatocyte follistatin secretion is regulated by the GCKR-GCK complex. [Diagram 2] Clustering of diabetes progression cohorts. [Figure 3A] Importance of five variables (plasma follistatin, proinsulin, insulin, C-peptide, baseline HbA1c) and variable selection (Figures 3A, 3B, 3C). [Figure 3B] Importance of five variables (plasma follistatin, proinsulin, insulin, C-peptide, baseline HbA1c) and variable selection (Figures 3A, 3B, 3C). [Figure 3C] Importance of five variables (plasma follistatin, proinsulin, insulin, C-peptide, baseline HbA1c) and variable selection (Figures 3A, 3B, 3C). [Figure 4] Performance and validation of four models for assessing the 4-year risk of developing type 2 diabetes in this cohort. [Diagram 5] ROC curve of the selection model (NNET) with four biomarkers with 10-fold cross-validation. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] Table 1: ROC AUC for 10-fold cross-validation on different variables with different methods. Table 2: Performance of each cluster by 10-fold cross-validation. Table 3: Comparison of AUC between models with and without follistatin for each class.

[0008] Detailed Description The present invention is based on the surprising realization by the inventors that follistatin is a biomarker for the early, high risk, onset of type 2 diabetes in humans.

[0009] The present invention thus relates to diagnosis and / or prediction for individuals who are at high risk of developing type 2 diabetes within a short period of time, such as less than 10 years, for example less than 9 years, less than 8 years, less than 7 years, less than 6 years, less than 5 years or less than 4 years, if that individual's condition is left untreated. A particular advantage of early diagnosis and / or prediction is the ability to prevent the onset of the disease by prophylactic treatment.

[0010] Thus, a first embodiment and aspect of the present invention details the use of follistatin as a biomarker in a method for the diagnosis of short-term, high-risk, onset of type 2 diabetes in humans.

[0011] Another aspect of the first embodiment and aspect of the present invention details the use of the biomarker, follistatin, in a method for predicting short-term, high-risk, onset of type 2 diabetes in humans.

[0012] In one embodiment of the first aspect, a method for diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human is detailed using follistatin according to a first embodiment, such as high-risk onset of type 2 diabetes in less than 10 years, preferably less than 9 years, less than 8 years, less than 7 years, less than 6 years, less than 5 years or less than 4 years.

[0013] In one embodiment of the first aspect, the method for diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human comprises determining baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c At least one further biomarker selected from the group consisting of baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c In one embodiment, the present invention details the use of follistatin according to any of the above embodiments, comprising k-means clustering to assess the level of risk of progressing to type 2 diabetes using at least two, at least three, or at least four additional biomarkers selected from:

[0014] In one embodiment of the first aspect, a method for diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human is detailed using follistatin according to any of the above embodiments, comprising assessing available biomarkers by recursive feature reduction for establishing a risk prediction model.

[0015] In one embodiment of the first aspect, a method for diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human comprises measuring follistatin and baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c and comparing the measured blood levels to model values ​​based on mean blood level values ​​from a group of humans at high risk of developing type 2 diabetes within less than 10 years, preferably within less than 5 years, or more preferably within less than 4 years.

[0016] In one embodiment of the first aspect, the method for diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human comprises determining baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c

[0033] This invention details the use of follistatin according to any of the above embodiments, comprising measuring blood levels of at least two, at least three, or at least four additional biomarkers selected from:

[0017] In one embodiment of the first aspect, the method of diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human is a method of predicting short-term, high-risk, onset of type 2 diabetes in a human, using follistatin according to any of the above embodiments.

[0018] In a second aspect of the present invention, follistatin and baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c and comparing the measured blood levels with model values ​​based on mean blood level values ​​from a group of humans at high risk of developing type 2 diabetes within less than 10 years, preferably within less than 5 years or more preferably within less than 4 years.

[0019] In one embodiment of the second aspect, the baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c In one embodiment, blood levels of at least two, at least three, or at least four additional biomarkers selected from are measured.

[0020] Similarly, follistatin is detailed herein for use as a biomarker in the diagnosis and / or prediction of short-term, high-risk, and onset of type 2 diabetes in humans.

[0021] In one embodiment thereof, follistatin is detailed for use as a biomarker in diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in a human, where the short-term, high-risk, onset of type 2 diabetes in a human is a high-risk onset of type 2 diabetes in less than 10 years, preferably less than 9 years, less than 8 years, less than 7 years, less than 6 years, less than 5 years, or less than 4 years.

[0022] In one embodiment, the diagnosis of short-term, high-risk, onset of type 2 diabetes in a human is based on baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c Preferably, baseline HbA was measured using at least one additional biomarker selected from 1c , proinsulin, C-peptide or 48-month HbA 1c The present invention details follistatin for use as a biomarker in diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in humans, including k-means clustering to assess the level of risk of progression to type 2 diabetes using at least two, at least three, or at least four additional biomarkers selected from:

[0023] In one embodiment thereof, the present invention details follistatin for use as a biomarker in diagnosing and / or predicting short-term, high-risk, onset of type 2 diabetes in humans, where the diagnosis of short-term, high-risk, onset of type 2 diabetes in humans comprises assessing available biomarkers by recursive feature reduction in a risk prediction model.

[0024] In one embodiment, the method comprises constructing a biomarker signature for early prediction of type 2 diabetes in humans, comprising measuring follistatin and baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c and comparing the measured blood levels with model values ​​based on mean blood level values ​​from a group of humans at high risk of developing type 2 diabetes in less than 10 years, preferably less than 5 years or more preferably less than 4 years.

[0025] In one embodiment, the baseline HbA 1c , proinsulin, C-peptide or 48-month HbA 1c The present invention details follistatin for use as a biomarker in diagnosing and / or predicting short-term, high risk, onset of type 2 diabetes in humans, further comprising measuring blood levels of at least two, at least three, or at least four additional biomarkers selected from:

[0026] The present invention relates to the diagnosis and identification of individuals who are at high risk of developing type 2 diabetes within a short period of time, such as less than 10 years, for example less than 9 years, less than 8 years, less than 7 years, less than 6 years, less than 5 years or less than 4 years, if their condition is left untreated. A particular advantage of early diagnosis is the ability to prevent the onset of the disease by prophylactic treatment.

[0027] According to the invention, the invention in an embodiment relates to diagnosing and / or predicting the risk of developing type 2 diabetes in a person within less than 10 years, such as within less than 9 years, less than 8 years, less than 7 years, less than 6 years, less than 5 years or less than 4 years, if the person's condition is left untreated.

[0028] In one embodiment, if an individual exhibits at least 2000 pg / mL of follistatin in serum, there is an increased risk of developing type 2 diabetes within a short period of time.

[0029] In one embodiment thereof, if an individual further presents at least 20 pmol / L of proinsulin in serum, there is an increased risk of developing type 2 diabetes within a short period of time.

[0030] In one embodiment thereof, if an individual further exhibits at least 5 ng / mL of C-peptide in serum, there is an increased risk of developing type 2 diabetes within a short period of time.

[0031] In one embodiment thereof, if the individual further exhibits at least 800 pg / mL of insulin in serum, there is an increased risk of developing type 2 diabetes within a short period of time.

[0032] As described in the Examples below, fasting individuals with serum levels of follistatin, or follistatin and one or more of the above biological markers, as provided above and in the Examples, may have a decreased HbA 1c and 48-month HbA 1c When measured using the markers, a statistically significant incidence of type 2 diabetes was observed over a control population including non-progressive pre-diabetic and non-diabetic individuals, and thus a population at high risk of developing type 2 diabetes. By themselves, with the notable exception of follistatin, the markers were not statistically distinguishable between the populations, but the combination of markers provided a clear identification. In the case of follistatin, serum levels above 2500 pg / mL, preferably above 3000 pg / mL, were significant by themselves in indicating a high risk of developing type 2 diabetes in the individuals studied. EXAMPLES

[0033] Working Example Reported study objectives The aim of this study was to develop a predictive model for assessing type 2 diabetes (T2D) risk through blood biomarker signatures.

[0034] Research design and methods The subjects were from a longitudinal cohort including 152 non-diabetic participants followed for 4 years for T2D progression. 1c , proinsulin, C-peptide, follistatin and 48-month HbA 1c The cohort was divided into clusters by k-means to assess the T2D progression risk level using ANOVA. Available biomarkers were assessed by recursive feature reduction to establish a risk prediction model. T2D 4-year prediction based on risk clustering was tested by neural network (NNET), support vector machine (SVM), random forest (RF) and generalized logistic regression (GLM) machine learning methods. The performance of four candidate risk models was assessed using 10-fold cross-validation.

[0035] Overall Results The cohort was divided into three risk groups: high risk, intermediate risk, and low risk. After biomarker validation, baseline HbA 1c , proinsulin, C-peptide and follistatin were selected. Using these four biomarkers, an optimal model was developed to assess an individual's 4-year risk of developing T2D by the NNET period learning method. The areas under the curve (AUC) of the receiver operating characteristic (ROC) curves for prediction of each risk group are 0.9 (high risk), 0.96 (intermediate risk) and 0.99 (low risk), respectively (10-fold cross-validation). The average AUC of the three risk groups is 0.97.

[0036] Therefore, in this specification, HbA 1cUsing a model incorporating four blood biomarkers, including proinsulin, C-peptide and follistatin, a novel tool for assessing type 2 diabetes risk four years before disease onset is presented.

[0037] method Cohort Participants The cohort was a multicenter, randomized, double-blind, placebo-controlled, clinical trial recruited in the United States. 1c Patient HbA from baseline over the 4-year study period was measured and concomitant medications were recorded in patients at baseline, 1 year, 2 years, and 4 years. 1c From the cohort, a T2D progression subcohort of approximately 400 patients was selected based on changes in blood glucose level and the absence of antidiabetic medication in these patients.

[0038] Plasma protein biomarker measurements Fasting insulin, Pro-insulin, C-peptide and follistatin were measured by enzyme-linked immunosorbent assay (ELISA) in baseline EDTA-plasma samples from 314 patients selected for inclusion in the type 2 diabetes progression cohort. For all assays except C-peptide, samples were measured in technical replicates and subsequent intra-assay mean coefficients of variation (%CV) were determined. Inter-assay %CV was calculated using internal controls. Internally optimized protocol [3~5]Insulin concentrations were determined using a custom electrochemiluminescence immunoassay and a MESO QuickPlex SQ 120 (Meso Scale Discovery (MSD), Gaithersburg, MD) according to the method described above. For the insulin assay, the intra-assay CV was 10.9% and the inter-assay CV was 3.1%. C-peptide levels were quantified with an electrochemiluminescence immunoassay using a Cobas e411 (Roche Diagnostics, Mannheim, Germany). Intact proinsulin was measured after 4-fold dilution in sample buffer using a colorimetric ELISA, calibrated against the WHO 1st International Standard for proinsulin, according to the manufacturer's instructions (IRP 84 / 611; catalog #IV2-102E, Immuno-Biological Laboratories, Inc., Minneapolis, MN). Assay absorbance was measured using a PHERAstar FSX (BMG Labtech Inc., Cary, NC). For the proinsulin assay, intra-assay CV was 6.6% and inter-assay CV was 10%. Plasma follistatin levels were measured after 2-fold dilution in sample diluent using a colorimetric ELISA according to the manufacturer's instructions (catalog #DFN00, R&D Systems, Minneapolis, MN). Assay absorbance was measured using a PHERAstar FSX (BMG Labtech Inc., Cary, NC). For the follistatin assay, intra-assay CV was 2.1% and inter-assay CV was 10%.

[0039] The model development process: Using machine learning to derive 4-year type 2 diabetes risk All statistical analyses were performed using the statistical analysis software package, Machine Learning Toolkit (Caret package). [6] ) and the statistical computing environment R [7] The significance of the results was set at p≦0.05.

[0040] To identify those blood parameters with the best predictive performance, feature selection was performed by recursive feature reduction implemented in the machine learning R package Caret (e.g., rfe, rfefilter). Five candidate biomarkers were evaluated for inclusion in the multimarker model. Since predictive techniques may perform differently on the data, the inventors used the R environment [7] Within the Caret package [6] Using the method, we chose to use four different types of predictive models. Four machine approaches were incorporated, including NNET (neural network), SVM (support vector machine), RF (random forest) and GLM (generalized logistic regression). Multivariate data analysis (NNET, SVM, RF and GLM) was used to investigate whether a blood-based biomarker panel allows risk prediction of developing type 2 diabetes. To ensure the robustness of the results, the best performing NNET model was evaluated in a 10-fold cross-validation procedure. The following characteristic quantities were calculated: AUC, accuracy, sensitivity and specificity.

[0041] Finally, receiver operating characteristic (ROC) curves were generated. The R library pROC for the two correlated ROC curves (i.e., ROC curves for high risk with NNET and NNET without follistatin) [8] The DeLong test was performed using the receiver operating characteristic (ROC) test. A P value <0.05 was considered significant.

[0042] The results indicate that the short-term, high-risk predicted outcomes are model-independent and therefore reflect the actual biological processes underlying the statistics of the groups studied.

[0043] result To identify genetic factors that affect plasma follistatin levels, the inventors performed GWAS in two additional different cohorts. Glucokinase Regulatory Protein (GCKR) was identified as a genetic regulator of plasma follistatin levels. Here, the inventors show that GCKR regulates hepatic follistatin secretion together with glucagon and insulin in human hepatocyte cell line HepG2. Previous studies have shown that GCKR forms a tight complex with GCK in the nucleus, and dissociation of GCK-GCKR binding leads to increased GCK translocation from the nucleus to the cytoplasm, which regulates hepatic cell glucose uptake and glycolysis.

[0044] In this study, HepG2 cells were transfected with GCK or co-transfected with GCK and GCKR expression plasmids (1:3 molar ratio). Additionally, cells were treated with and without AMG-3969, a GCK-GCKR complex disruptor that promotes robust translocation of dissociated GCK from the nucleus to the cytoplasm. Cells were incubated with glucagon (1 μg / ml) and the intracellular cAMP activator forskolin (20 μM) in low glucose DMEM medium (5.5 mM), conditions previously shown to stimulate follistatin secretion in hepatocytes.

[11] In the presence of the GCK-GCKR complex and its disruptant AMG-3969, follistatin secretion was increased by 40% compared to controls (Figure 4A), which was reversed by co-incubation with insulin (Figure 4B). Transfection with GCK alone or co-transfection with GCK-GCKR without AMG-3969 (which does not affect translocation of GCKR from the nucleus to the cytoplasm) had no effect on follistatin secretion (Figure 1).

[0045] Figure 1. Hepatocyte follistatin secretion is controlled by the GCKR-GCK complex. Human hepatocarcinoma-derived HepG2 cells were transfected with the indicated plasmids: i) control (pCMV-XL4, white bars); ii) GCK:GCKR (1:0; no GCKR, grey bars); iii) GCK:GCKR (1:3, black bars). 48 hours after transfection, cells were serum-starved in low glucose (5.5 mM) DMEM for 3 hours and the GCKR-GCK disruptor AMG-3969 (0.7 μM) was added in this medium for 30 minutes. Cells were then incubated in serum-free low glucose (5.5 mM) DMEM containing glucagon (1 μg / ml) and forskolin (20 μM) and AMG-3969 (0.7 μM) was added to the corresponding wells. After 4 hours of incubation, the medium was collected for follistatin assay by ELISA. Follistatin levels were normalized to protein concentration within each sample. Two independent experiments with three technical replicates per condition were performed on different days using different plasmid preparations and cell passage numbers. B. HepG2 cells were treated as described in panel A, but in the presence of insulin (100 nM). As shown, * p<0.05 and ** p<0.01.

[0046] To better characterize type 2 diabetes risk among individuals without diabetes, US cohort participants were divided into clusters using follistatin and other variables previously shown to be associated with diabetes or future diabetes risk. 1c , proinsulin, C-peptide, follistatin and 48-month HbA 1c K-means clustering using β-category 1 (β-category 1) identified three risk groups: high, intermediate, and low risk (Figure 2). The high-risk group in cluster 1 progressed from non-diabetic to diabetes after 48 months, with a median HbA of 5.6% at baseline to 6.8% at 48 months. 1c The intermediate risk group in cluster 2 was non-progressive prediabetes (median HbA 1c HbA 1c48 months (6.3%), and low-risk group 3 of cluster 3 included non-progressing non-diabetic individuals (median HbA 1c HbA 1c Patients from cluster 1 had significantly higher plasma follistatin levels at baseline (Figure 2B), as well as higher baseline plasma proinsulin (Figure 2C), C-peptide (Figure 2D), and insulin levels (Figure 2E) than the other clusters 48 months before diabetes onset.

[0047] Figure 2. Diabetes progression cohort clustering. Baseline HbA 1c , plasma follistatin, proinsulin, C-peptide and HbA 1c Individuals (n=152) from the US cohort were clustered by unsupervised K-means using . A. Cluster 1: Non-diabetic to diabetes progression (white bars; median HbA 1c HbA 1c 48 months 6.8%; n=20); Cluster 2: Prediabetes non-progression (gray bar; median HbA 1c HbA 1c 48 months 6.3%; n=62); Cluster 3: Non-diabetic non-progression (black bars; median HbA1c baseline 5.4% to HbA 1c 48 months 5.5%; n=70). B–E. Cluster distribution plots of baseline follistatin (pg / mL, B), proinsulin (pmol / L, C), C-peptide (ng / mL, D) and insulin (pg / mL, E). BL: baseline; 48M: HbA 1c 48 months. In the figure, **** p<0.0001, *** p<0.001, ** p<0.01, * p<0.05.

[0048] Each baseline variable (HbA 1cTo validate the predictive power of the three markers (i.e., proinsulin, C-peptide, insulin and follistatin) and their combinations, the inventors analyzed the data using neural network (NNET), support vector machine (SVM), random forest (RF) and generalized logistic regression (GLM) machine learning methods. [6] Recursive feature reduction was performed using the rfe-function of (Table 1). For the 4-year type 2 diabetes prediction for each risk group, proinsulin and follistatin were significantly associated with HbA 1c and follistatin have the greatest importance for the medium and low risk groups (Figure 3B). The overall importance for the three risk groups is shown by the maximum operation (Figure 3C). The combination of the four variables gives the greatest ref accuracy (10-fold cross-validation, Figure 3A). Finally, the four top risk factors (baseline HbA 1c , follistatin, proinsulin and C-peptide) were selected.

[0049] Figure 3. Five variables (plasma follistatin, proinsulin, insulin, C-peptide, baseline HbA1 c ) and the importance of variable selection. Accuracy with different variables using recursive feature reduction (3A), involvement of each variable at different risk levels (3B) and maximum score of three risk levels for each variable (3C).

[0050] Next, different machine learning methods were compared to test the predictive performance of 4-year type 2 diabetes risk incorporating the selected biomarkers. NNET, SVM, RF and GLM machine learning methods were evaluated by 10-fold cross-validation (Table 1). The sensitivity, specificity, accuracy and AUC ranges (Figure 4A) and confidence intervals (Figure 4B) were compared among the four models. NNET remained stable and performed substantially better than the other three models in sensitivity and specificity. The mean values ​​of sensitivity and specificity were greater than 0.84 (Table 2). Furthermore, in the comparison of accuracy and AUC among the four models, NNET obtained a higher performance. The receiver operating characteristic (ROC) curve and area under the curve (AUC) were 0.9 for the high-risk group, 0.96 for the medium-risk group, and 0.99 for the low-risk group, respectively (10-fold cross-validation). Adding follistatin to the NNET model significantly improved the AUC in the high-risk group (AUC 0.9 <with follistatin> vs. 0.75 <without follistatin>, P = 4e-04 DeLong test). In the medium-risk group, the AUC improved (AUC 0.99 <with follistatin> vs. 0.96 <without follistatin>, P = le-02 DeLong test), while in the low-risk group (AUC 0.96 <with follistatin> vs. 0.95 <without follistatin>, P = le-01 DeLong test) (Table 3 and Figure 5).

[0051] Figure 4. Performance and validation of four models for assessing the 4-year risk of developing type 2 diabetes in this cohort. Sensitivity, specificity, accuracy and range of AUC (4A) and confidence levels (4B) are shown for the four models.

[0052] Figure 5. ROC curves for the 10-fold cross-validated four-biomarker selection model (NNET). The ROC curves for the four biomarkers (baseline HbA 1cFigure 1 shows the ROC curves for the NNET model incorporating NNET signatures (, follistatin, proinsulin and C-peptide) (10-fold cross-validation). The DeLong test for the ROC curves of the signatures with and without follistatin is p=4e-04 for the high risk group, p=0.01 for the medium risk group and p=0.1 for the low risk group.

[0053] Consideration Based on the results of the "risk clustering analysis," the inventors used four variables to construct a biomarker signature for predicting future diabetes: baseline follistatin, HbA 1c , proinsulin and C-peptide.

[0054] A multi-biomarker model was developed to assess risk of type 2 diabetes by four blood b biomarkers using multiple statistical approaches. The performance of the NNET model is better than any other baseline measure of risk. The NNET model offers a more convenient alternative for obtaining risk estimates: the laboratory measures biomarker concentrations in fasting blood samples and returns a calculated risk level. The NNET model does not rely on anthropometric measurements or self-reported risk factors (such as family history or smoking).

[0055] The four biomarkers selected for the NNET model are involved in various biological pathways. Proinsulin is an important indicator of metabolic disorders, including diabetes and obesity. It has been shown that disproportionate secretion of proinsulin, an insulin precursor, may not only be a specific indicator of insulin resistance but also a hallmark of β-cell dysfunction. [9] Follistatin is a secreted protein expressed in almost all major tissues, and studies have demonstrated its association with metabolic disease. [10、11] , and plasma levels have been suggested to be elevated in patients with type 2 diabetes.

[10] Circulating follistatin has a direct effect on glucose metabolism in humans by increasing insulin and inhibiting glucagon secretion from the pancreas.

[12] However, as disclosed in this disclosure, it was previously unknown whether follistatin predicts the onset of type 2 diabetes prior to the onset of type 2 diabetes.

[0056] Localized overexpression of follistatin in the pancreas of diabetic mice resulted in elevated serum insulin levels

[13] Recent studies by Tao et al. have identified follistatin as a mediator of systemic metabolic dysregulation associated with diabetes.

[14] In hyperglycemic and obese mice fed a high-fat diet, knockdown of follistatin restored glucose tolerance, white adipose tissue insulin signaling, and insulin-induced suppression of hepatic glucose production. As demonstrated in this disclosure, it was not previously known that hepatic follistatin secretion is regulated by GCKR together with glucagon and insulin (Figure 1).

[0057] In obese individuals with diabetes who underwent gastric bypass surgery, serum follistatin was significantly increased in HbA 1c The HbA 1c is primarily measured to determine the 3-month average plasma glucose concentration and therefore can be used as a diagnostic test for diabetes.

[0058] A significant association was found between serum C-peptide levels and the risk of diabetes and prediabetes among Chinese women with a history of gestational diabetes.

[0059] Previous findings have suggested that elevated C-peptide levels may be a predictor of diabetes and prediabetes.

[15] .

[0060] Several mathematical models using machine learning methods were run on the variables: SVM, NNET, RF and GLM. Among all the tested methods, NNET gave the best performance. Using the selected biomarker signature, NNET predicts with very high specificity and sensitivity whether an individual has low, medium or high risk of developing diabetes within 4 years. The AUC is 0.9 (10-fold cross-validation) for predicting high risk, 0.99 for predicting medium risk and 0.96 for predicting low risk (Figure 3). AUC comparison between these models with and without follistatin showed that the multiple biomarkers performed better than the single biomarkers and without follistatin.

[0061] In summary, by applying various statistical methods for biomarker selection, we developed an NNET model incorporating up to four circulating biomarkers. The NNET provides superior assessment of diabetes risk compared to models with only a single biomarker and without follistatin. Our results suggest that the NNET model may be an important tool to identify individuals at highest risk of developing type 2 diabetes, a population for which the most comprehensive prevention strategies should be considered. The improved performance of this model compared to that of a single marker highlights the value of risk assessment models incorporating multiple biomarkers, including follistatin, from a wide variety of pathophysiological pathways related to type 2 diabetes.

[0062] References [1] Bommer, C. et al. Global Economic Burden of Diabetes in Adults: Projections from 2015 to 2030. Diabetes Care 41,963-970,doi:10.2337 / dcl7-1962(2018). [2]Knowler,W.C.et al.Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin.N Engl J Med 346,393-403,doi:10.1056 / NEJMoa012512(2002). [3]Lin,H.V.et al.GPR142 Controls Tryptophan-Induced Insulin and Incretin Hormone Secretion to Improve Glucose Metabolism.PLoS One 11,e0157298,doi:10.1371 / journal.pone.0157298(2016). [4]Bueno,A.B.et al.Positive Allosteric Modulation of the Glucagon-like Peptide-1 Receptor by Diverse Electrophiles.J Biol Chem 291,10700-10715,doi:10.1074 / jbc.M115.696039(2016). [5]Farb,T.B.et al.Regulation of Endogenous(Male)Rodent GLP-1 Secretion and Human Islet Insulin Secretion by Antagonism of Somatostatin Receptor 5.Endocrinology 158,3859-3873,doi:10.1210 / en.2017-00639(2017). [6]Kuhn,M.Caret:classification and regression training.Astrophysics Source Code Library(2015). [7]Team,R.C.R:A Language and Environment for Statistical Computing,dim(ca533)1,34(2018). [8]Robin,X.et al.pROC:an open-source package for R and S+ to analyze and compare ROC curves.BMC bioinformatics 12,77(2011). [9]Russo,G.T.et al.Factors associated with beta-cell dysfunction in type 2 diabetes:the BETADECLINE study.PLoS One 9,e109702(2014).

[10] Hansen,J.et al.Plasma follistatin is elevated in patients with type 2 diabetes:relationship to hyperglycemia,hyperinsulinemia,and systemic low-grade inflammation.Diabetes Metab Res Rev 29,463-472,doi:10.1002 / dmrr.2415(2013).

[11] Yndestad,A.et al.A complex role of activin A in non alcoholic fatty liver disease.Am J Gastroenterol 104,2196-2205,doi:10.1038 / ajg.2009.318(2009).

[12] Hansen,J.S.et al.Circulating Follistatin Is Liver-Derived and Regulated by the Glucagon-to-Insulin Ratio.J Clin Endocrinol Metab 101,550-560,doi:10.1210 / jc.2015-3668(2016).

[13] Zhao, C. et al. Overcoming Insulin Insufficiency by Forced Follistatin Expression in beta-cells of db / db Mice.Mol.Ther.23,866-874,doi:10.1038 / mt.2015.29(2015).

[14] Tao, R. et al. Inactivating hepatic follistatins alleviating hyperglycemia.Nat Med 24,1058-1069,doi:10.1038 / s41591-018-0048-0(2018).

[15] Yin, P. et al.C-peptide levels and the risk of diabetes and pre-diabetes among Chinese women with gestational diabetes.Journal of diabetes and its complications 31,1658-1662(2017).

[0063] Conclusion The term "comprising", when used in the claims, does not exclude other elements or steps. The terms "a" or "an", when used in the claims, do not exclude a plurality. Although the invention has been described in detail for purposes of illustration, it will be understood that such details are for the purposes only and that modifications thereof can be made by those skilled in the art without departing from the scope of the invention.

[0064] [Table 1]

[0065] [Table 2]

[0066]

Table 3

Claims

1. Baseline HbA １ｃ , proinsulin, C-peptide or 48-month HbA １ｃ 2. Use of follistatin as a blood biomarker in a method for the diagnosis and / or prediction of short-term, high risk, or onset of type 2 diabetes in a human, in combination with at least three of the following:

2. 2. The use of follistatin according to claim 1, wherein the method for diagnosis and / or prediction of short-term, high risk or onset of type 2 diabetes in humans comprises k-means clustering to assess the level of risk of progressing to type 2 diabetes.

3. 3. Use of follistatin according to claim 1 or 2, wherein the method for diagnosis and / or prediction of short-term, high risk or onset of type 2 diabetes in humans comprises assessing available biomarkers by recursive feature reduction to establish a risk prediction model.

4. The method for diagnosing and / or predicting short-term, high risk, or onset of type 2 diabetes in a human comprises measuring follistatin and baseline HbA １ｃ , proinsulin, C-peptide or 48-month HbA １ｃ and comparing said measured blood levels with model values ​​based on mean blood level values ​​from a group of humans at high risk of developing type 2 diabetes within less than four years.

5. 5. The use of follistatin according to any of claims 1 to 4, wherein the method for diagnosing and / or predicting short-term, high risk or onset of type 2 diabetes in a human is a method for diagnosing short-term, high risk or onset of type 2 diabetes in a human.

6. 6. The use of follistatin according to any one of claims 1 to 5, wherein the method for diagnosis and / or prediction of short-term, high risk or onset of type 2 diabetes in a human is a method for predicting short-term, high risk or onset of type 2 diabetes in a human.

7. Follistatin and baseline HbA １ｃ , proinsulin, C-peptide or 48-month HbA １ｃ and comparing said measured blood levels with model values ​​based on mean blood level values ​​from a group of humans at high risk of developing type 2 diabetes within less than four years.

8. Baseline HbA １ｃ , proinsulin, C-peptide or 48-month HbA １ｃ 8. A method for constructing a biomarker signature for early prediction of type 2 diabetes in humans as described in claim 7, comprising measuring blood levels of at least three or at least four additional biomarkers selected from:

Images