From heterogeneous healthcare data to disease-specific biomarker networks: A hierarchical Bayesian network approach

In this work, we introduce an entirely data-driven and automated approach to reveal disease-associated biomarker and risk factor networks from heterogeneous and high-dimensional healthcare data. Our workflow is based on Bayesian networks, which are a popular tool for analyzing the interplay of biomarkers. Usually, data require extensive manual preprocessing and dimension reduction to allow for effective learning of Bayesian networks. For heterogeneous data, this preprocessing is hard to automatize and typically requires domain-specific prior knowledge. We here combine Bayesian network learning with hierarchical variable clustering in order to detect groups of similar features and learn interactions between them entirely automated. We present an optimization algorithm for the adaptive refinement of such group Bayesian networks to account for a specific target variable, like a disease. The combination of Bayesian networks, clustering, and refinement yields low-dimensional but disease-specific interaction networks. These networks provide easily interpretable, yet accurate models of biomarker interdependencies. We test our method extensively on simulated data, as well as on data from the Study of Health in Pomerania (SHIP-TREND), and demonstrate its effectiveness using non-alcoholic fatty liver disease and hypertension as examples. We show that the group network models outperform available biomarker scores, while at the same time, they provide an easily interpretable interaction network.     

Absence of a memory effect for the insulinotropic action of glucagon-like peptide 1 (GLP-1) in healthy volunteers.

BACKGROUND/AIMS: The term memory effect refers to the phenomenon that B cell stimuli retain some of their insulinotropic effects after they have been removed. Memory effects exist for glucose and sulfonylureas. It is not known whether there is a B-cell memory for incretin hormones such as GLP-1. SUBJECTS/METHODS: Eight healthy young volunteers were studied on four occasions in the fasting state. In one experiment, placebo was administered (a). in three more experiments (random order), synthetic GLP-1 (7 - 36 amide) at 1.2 pmol/kg/min was administered over a period of three hours. At 0 min, a bolus of glucose was injected intravenously (0.33 g/kg body weight). GLP-1 was infused from (b). - 60 to 120 min, (c). - 210 to - 30 min, or (d). - 300 to - 120 min. Glucose (glucose oxidase), insulin, C-peptide, GLP-1, and glucagon (immunoassays) were determined. Statistical analysis was carried out by ANOVA and appropriate post hoc tests. RESULTS: GLP-1 plasma levels during the infusion periods were elevated to 89 +/- 9, 85 +/- 13, and 89 +/- 6 pmol/l (p < 0.0001 vs. placebo, 10 +/- 1 pmol/l). Glucose was eliminated faster (p < 0.0001), with an enhanced negative rebound (p = 0.014), and insulin and C-peptide increments were greater after intravenous glucose administration (p < 0.0001) if GLP-1 was administered during the injection of the glucose bolus, but not if GLP-1 had been administered until 120 or 30 min before the glucose load. There was a trend towards higher insulin concentrations (p = 0.056) five minutes after glucose with GLP-1 administered until - 30 min before the glucose load. Glucagon was suppressed by exogenous glucose, but increased significantly (p = 0.013) during the induction of reactive hypoglycemia after glucose injection during GLP-1 administration. CONCLUSION: 1). No memory effect appears to exist for insulinotropic actions of GLP-1, in line with clinical data. 2). Reactive hypoglycemia causes a prompt rise in glucagon despite pharmacological circulating concentrations of GLP-1. 3). Similar studies should be performed in Type 2-diabetic patients, because exposure to GLP-1 might recruit dormant pancreatic B cells to become glucose-competent, and this might contribute to the overall antidiabetogenic effect of GLP-1 in such patients.     

The glucagon-like peptide-1 metabolite GLP-1-(9-36) amide reduces postprandial glycemia independently of gastric emptying and insulin secretion in humans

Glucagon-like peptide 1 (GLP-1) lowers glycemia by modulating gastric emptying and endocrine pancreatic secretion. Rapidly after its secretion, GLP-1-(7-36) amide is degraded to the metabolite GLP-1-(9-36) amide. The effects of GLP-1-(9-36) amide in humans are less well characterized. Fourteen healthy volunteers were studied with intravenous infusion of GLP-1-(7-36) amide, GLP-1-(9-36) amide, or placebo over 390 min. After 30 min, a solid test meal was served, and gastric emptying was assessed. Blood was drawn for GLP-1 (total and intact), glucose, insulin, C-peptide, and glucagon measurements. Administration of GLP-1-(7-36) amide and GLP-1-(9-36) amide significantly raised total GLP-1 plasma levels. Plasma concentrations of intact GLP-1 increased to 21 +/- 5 pmol/l during the infusion of GLP-1-(7-36) amide but remained unchanged during GLP-1-(9-36) amide infusion [5 +/- 3 pmol/l; P < 0.001 vs. GLP-1-(7-36) amide administration]. GLP-1-(7-36) amide reduced fasting and postprandial glucose concentrations (P < 0.001) and delayed gastric emptying (P < 0.001). The GLP-1 metabolite had no influence on insulin or C-peptide concentrations. Glucagon levels were lowered by GLP-1-(7-36) amide but not by GLP-1-(9-36) amide. However, the postprandial rise in glycemia was reduced significantly (by approximately 6 mg/dl) by GLP-1-(9-36) amide (P < 0.05). In contrast, gastric emptying was completely unaffected by the GLP-1 metabolite. The GLP-1 metabolite lowers postprandial glycemia independently of changes in insulin and glucagon secretion or in the rate of gastric emptying. Most likely, this is because of direct effects on glucose disposal. However, the glucose-lowering potential of GLP-1-(9-36) amide appears to be small compared with that of intact GLP-1-(7-36) amide.     

Glucagon-like peptide 1 and its derivatives in the treatment of diabetes

Glucagon-like peptide 1 (GLP-1) was discovered as an insulinotropic gut hormone, suggesting a physiological role as an incretin hormone, i.e., being responsible, in part, for the higher insulin secretory response after oral as compared to intravenous glucose administration. This difference, the incretin effect, is partially lost in patients with Type 2 diabetes. The actions of GLP-1 include (a) a stimulation of insulin secretion in a glucose-dependent manner, (b) a suppression of glucagon, (c) a reduction in appetite and food intake, (d) a deceleration of gastric emptying, (e) a stimulation of beta-cell neogenesis, growth and differentiation in animal and tissue culture experiments, and (f) an in vitro inhibition of beta-cell apoptosis induced by different toxins. Intravenous GLP-1 can normalize and subcutaneous GLP-1 can significantly lower plasma glucose in the majority of patients with Type 2 diabetes. GLP-1 itself, however, is inactivated rapidly in vivo and thus does not appear to be useful as a therapeutic agent in the long-term treatment of Type 2 diabetes. Other agents acting on GLP-1 receptors have been found (like exendin-4) or developed as GLP-1 derivatives (like liraglutide or GLP-1/CJC-1131). Clinical trials with exenatide (two injections per day) and liraglutide (one injection per day) have shown reductions in glucose concentrations and HbA1c by more than 1%, associated with moderate weight loss (2-3 kg), but also some nausea and, rarely, vomiting. It is hoped that this new class of drugs interacting with the GLP-1 or other incretin receptors, the so-called "incretin mimetics", will broaden our armamentarium of antidiabetic medications in the nearest future.     

Gastric inhibitory polypeptide does not inhibit gastric emptying in humans

The insulinotropic gut hormone gastric inhibitory polypeptide (GIP) has been demonstrated to inhibit gastric acid secretion and was proposed to possess "enterogastrone" activity. GIP effects on gastric emptying have not yet been studied. Fifteen healthy male volunteers (23.9 +/- 3.3 yr, body mass index 23.7 +/- 2.3 kg/m(2)) were studied with the intravenous infusion of GIP (2 pmol.kg(-1).min(-1)) or placebo, each administered to the volunteers on separate occasions from -30 to 360 min in the fasting state. At 0 min, a solid test meal (250 kcal containing [(13)C]sodium octanoate) was served. Gastric emptying was calculated from the (13)CO(2) exhalation rates in breath samples collected over 360 min. Venous blood was drawn in 30-min intervals for the determination of glucose, insulin, C-peptide, and GIP (total and intact). Statistical calculations were made by use of repeated-measures ANOVA and one-way ANOVA. During the infusion, GIP rose to steady-state concentrations of 159 +/- 15 pmol/l for total and 34 +/- 4 pmol/l for intact GIP (P < 0.0001). Meal ingestion further increased GIP concentrations in both groups, reaching peak levels of 265 +/- 20 and 82 +/- 9 pmol/l for total and 67 +/- 7 and 31 +/- 9 pmol/l for intact GIP during the administration of GIP and placebo, respectively (P < 0.0001). There were no differences in glucose, insulin, and C-peptide between the experiments with the infusion of GIP or placebo. Gastric half-emptying times were 120 +/- 9 and 120 +/- 18 min (P = 1.0, with GIP and placebo, respectively). The time pattern of gastric emptying was similar in the two groups (P = 0.98). Endogenous GIP secretion, as derived from the incremental area under the curve of plasma GIP concentrations in the placebo experiments, did not correlate to gastric half-emptying times (r(2) = 0.15, P = 0.15 for intact GIP; r(2) = 0.21, P = 0.086 for total GIP). We conclude that gastric emptying does not appear to be influenced by GIP. The secretion of GIP after meal ingestion is not suppressed by its exogenous administration. The lack of effect of GIP on gastric emptying underlines the differences between GIP and the second incretin glucagon-like peptide 1.     

Reduction of hepatic insulin clearance after oral glucose ingestion is not mediated by glucagon-like peptide 1 or gastric inhibitory polypeptide in humans

Changes in hepatic insulin clearance can occur after oral glucose or meal ingestion. This has been attributed to the secretion and action of gastric inhibitory polypeptide (GIP) and glucagon-like peptide (GLP)-1. Given the recent availability of drugs based on incretin hormones, such clearance effects may be important for the future treatment of type 2 diabetes. Therefore, we determined insulin clearance in response to endogenously secreted and exogenously administered GIP and GLP-1. Insulin clearance was estimated from the molar C-peptide-to-insulin ratio calculated at basal conditions and from the respective areas under the curve after glucose, GIP, or GLP-1 administration. Oral glucose administration led to an approximately 60% reduction in the C-peptide-to-insulin ratio (P < 0.0001), whereas intravenous glucose administration had no effect (P = 0.09). The endogenous secretion of GIP or GLP-1 was unrelated to the changes in insulin clearance. The C-peptide-to-insulin ratio was unchanged after the intravenous administration of GIP or GLP-1 in the fasting state (P = 0.27 and P = 0.35, respectively). Likewise, infusing GLP-1 during a meal course did not alter insulin clearance (P = 0.87). An inverse nonlinear relationship was found between the C-peptide-to-insulin ratio and the integrated insulin levels after oral and during intravenous glucose administration. Insulin clearance is reduced by oral but not by intravenous glucose administration. Neither GIP nor GLP-1 has significant effects on insulin extraction. An inverse relationship between insulin concentrations and insulin clearance suggests that the secretion of insulin itself determines the rate of hepatic insulin clearance.     

Associations between thyroid hormones and serum metabolite profiles in an euthyroid population

Metabolomics - The aim was to characterise associations between circulating thyroid hormones—free thyroxine (FT4) and thyrotropin (TSH)—and the metabolite profiles in serum samples from...