Mahlet G. Tadesse

Bayesian Variable Selection in Clustering High-Dimensional Data

Over the last decade, technological advances have generated an explosion of data with substantially smaller sample size relative to the number of covariates (p ≫ n). A common goal in the analysis of such data involves uncovering the group structure of the observations and identifying the discriminating variables. In this article we propose a methodology for addressing these problems simultaneously. Given a set of variables, we formulate the clustering problem in terms of a multivariate normal mixture model with an unknown number of components and use the reversible-jump Markov chain Monte Carlo technique to define a sampler that moves between different dimensional spaces. We handle the problem of selecting a few predictors among the prohibitively vast number of variable subsets by introducing a binary exclusion/inclusion latent vector, which gets updated via stochastic search techniques. We specify conjugate priors and exploit the conjugacy by integrating out some of the parameters. We describe strategies for posterior inference and explore the performance of the methodology with simulated and real datasets.

Adipokines, insulin resistance, and coronary artery calcification.

OBJECTIVES We evaluated the hypothesis that plasma levels of adiponectin and leptin are independently but oppositely associated with coronary artery calcification (CAC), a measure of subclinical atherosclerosis. In addition, we assessed which biomarkers of adiposity and insulin resistance are the strongest predictors of CAC beyond traditional risk factors, metabolic syndrome, and plasma C-reactive protein (CRP). BACKGROUND Adipokines are fat-secreted biomolecules with pleiotropic actions that converge in diabetes and cardiovascular disease. METHODS We examined the association of plasma adipocytokines with CAC in 860 asymptomatic, nondiabetic participants in the SIRCA (Study of Inherited Risk of Coronary Atherosclerosis). RESULTS Plasma adiponectin and leptin levels had opposite and distinct associations with adiposity, insulin resistance, and inflammation. Plasma leptin was positively (top vs. bottom quartile) associated with higher CAC after adjustment for age, gender, traditional risk factors, and Framingham risk scores (tobit regression ratio 2.42 (95% confidence interval [CI]: 1.48 to 3.95; p = 0.002) and further adjustment for metabolic syndrome and CRP (tobit regression ratio: 2.31; 95% CI: 1.36 to 3.94; p = 0.002). In contrast, adiponectin levels were not associated with CAC. Comparative analyses suggested that levels of leptin, interleukin-6, and soluble tumor necrosis factor receptor-2, as well as the homeostasis model assessment of insulin resistance (HOMA-IR) index, predicted CAC scores, but only leptin and HOMA-IR provided value beyond risk factors, metabolic syndrome, and CRP. CONCLUSIONS In SIRCA, although both leptin and adiponectin levels were associated with metabolic and inflammatory markers, only leptin was a significant independent predictor of CAC. Of several metabolic markers, leptin and the HOMA-IR index had the most robust, independent associations with CAC.

The influence of pravastatin and atorvastatin on markers of oxidative stress in hypercholesterolemic humans.

OBJECTIVES The aim of this study was to determine the effects of pravastatin and atorvastatin on markers of oxidative stress in plasma. BACKGROUND Hydroxymethylglutaryl coenzyme A reductase inhibitors reduce low-density lipoprotein cholesterol (LDL-C) and cardiovascular risk, but their effects on circulating biomarkers of oxidative stress are not well-defined. METHODS Hypercholesterolemic subjects (n = 120, ages 21 to 80 years with LDL-C 130 to 220 mg/dl) were randomized in a double-blind, parallel design to pravastatin 40 mg/day (prava40), atorvastatin 10 mg/day (atorva10), atorvastatin 80 mg/day (atorva80), or placebo. At baseline and 16 weeks, urinary isoprostanes (8, 12-iso-iPF(2 alpha)-VI isoform), plasma lipoprotein-associated phospholipase A2 (Lp-PLA2), Mercodia oxidized LDL (OxLDL) with antibody 4E6, oxidized phospholipids/apolipoprotein B-100 particle (OxPL/apoB) with antibody E06, immunoglobulin (Ig)G/IgM autoantibodies to malondialdehyde (MDA)-LDL, and apolipoprotein B (apoB)-immune complexes (IC) were measured. RESULTS After 16 weeks, there were no significant changes in urinary 8, 12-iso-iPF(2 alpha)-VI. The Lp-PLA2 and OxLDL were reduced in statin-treated groups, but after adjusting for apoB, only prava40 led to a reduction in Lp-PLA2 (-15%, p = 0.008) and atorva10 to a decrease in OxLDL (-12.9%, p = 0.01). The OxPL/apoB increased 25.8% (p < 0.01) with prava40 and 20.2% (p < 0.05) with atorva80. There were no changes in MDA-LDL autoantibodies, but significant decreases in IC were noted. CONCLUSIONS This study suggests that statin therapy results in variable effects on oxidative stress markers in hypercholesterolemic subjects. Future outcome studies should collectively assess various oxidative markers to define clinical utility.

Utilization of metabolomics to identify serum biomarkers for hepatocellular carcinoma in patients with liver cirrhosis

Characterizing the metabolic changes pertaining to hepatocellular carcinoma (HCC) in patients with liver cirrhosis is believed to contribute towards early detection, treatment, and understanding of the molecular mechanisms of HCC. In this study, we compare metabolite levels in sera of 78 HCC cases with 184 cirrhotic controls by using ultra performance liquid chromatography coupled with a hybrid quadrupole time-of-flight mass spectrometry (UPLC-QTOF MS). Following data preprocessing, the most relevant ions in distinguishing HCC cases from patients with cirrhosis are selected by parametric and non-parametric statistical methods. Putative metabolite identifications for these ions are obtained through mass-based database search. Verification of the identities of selected metabolites is conducted by comparing their MS/MS fragmentation patterns and retention time with those from authentic compounds. Quantitation of these metabolites is performed in a subset of the serum samples (10 HCC and 10 cirrhosis) using isotope dilution by selected reaction monitoring (SRM) on triple quadrupole linear ion trap (QqQLIT) and triple quadrupole (QqQ) mass spectrometers. The results of this analysis confirm that metabolites involved in sphingolipid metabolism and phospholipid catabolism such as sphingosine-1-phosphate (S-1-P) and lysophosphatidylcholine (lysoPC 17:0) are up-regulated in sera of HCC vs. those with liver cirrhosis. Down-regulated metabolites include those involved in bile acid biosynthesis (specifically cholesterol metabolism) such as glycochenodeoxycholic acid 3-sulfate (3-sulfo-GCDCA), glycocholic acid (GCA), glycodeoxycholic acid (GDCA), taurocholic acid (TCA), and taurochenodeoxycholate (TCDCA). These results provide useful insights into HCC biomarker discovery utilizing metabolomics as an efficient and cost-effective platform. Our work shows that metabolomic profiling is a promising tool to identify candidate metabolic biomarkers for early detection of HCC cases in high risk population of cirrhotic patients.

INCORPORATING BIOLOGICAL INFORMATION INTO LINEAR MODELS: A BAYESIAN APPROACH TO THE SELECTION OF PATHWAYS AND GENES.

The vast amount of biological knowledge accumulated over the years has allowed researchers to identify various biochemical interactions and define different families of pathways. There is an increased interest in identifying pathways and pathway elements involved in particular biological processes. Drug discovery efforts, for example, are focused on identifying biomarkers as well as pathways related to a disease. We propose a Bayesian model that addresses this question by incorporating information on pathways and gene networks in the analysis of DNA microarray data. Such information is used to define pathway summaries, specify prior distributions, and structure the MCMC moves to fit the model. We illustrate the method with an application to gene expression data with censored survival outcomes. In addition to identifying markers that would have been missed otherwise and improving prediction accuracy, the integration of existing biological knowledge into the analysis provides a better understanding of underlying molecular processes.

LC-MS based serum metabolomics for identification of hepatocellular carcinoma biomarkers in Egyptian cohort.

Although hepatocellular carcinoma (HCC) has been subjected to continuous investigation and its symptoms are well-known, early stage diagnosis of this disease remains difficult and the survival rate after diagnosis is typically very low (3-5%). Early and accurate detection of metabolic changes in the sera of patients with liver cirrhosis can help improve the prognosis of HCC and lead to a better understanding of its mechanism at the molecular level, thus providing patients with in-time treatment of the disease. In this study, we compared metabolite levels in sera of 40 HCC patients and 49 cirrhosis patients from Egypt by using ultraperformance liquid chromatography coupled with quadrupole time-of-flight mass spectrometer (UPLC-QTOF MS). Following data preprocessing, the most relevant ions in distinguishing HCC cases from cirrhotic controls are selected by statistical methods. Putative metabolite identifications for these ions are obtained through mass-based database search. The identities of some of the putative identifications are verified by comparing their MS/MS fragmentation patterns and retention times with those from authentic compounds. Finally, the serum samples are reanalyzed for quantitation of selected metabolites as candidate biomarkers of HCC. This quantitation was performed using isotope dilution by selected reaction monitoring (SRM) on a triple quadrupole linear ion trap (QqQLIT) coupled to UPLC. Statistical analysis of the UPLC-QTOF data identified 274 monoisotopic ion masses with statistically significant differences in ion intensities between HCC cases and cirrhotic controls. Putative identifications were obtained for 158 ions by mass based search against databases. We verified the identities of selected putative identifications including glycholic acid (GCA), glycodeoxycholic acid (GDCA), 3β, 6β-dihydroxy-5β-cholan-24-oic acid, oleoyl carnitine, and Phe-Phe. SRM-based quantitation confirmed significant differences between HCC and cirrhotic controls in metabolite levels of bile acid metabolites, long chain carnitines and small peptide. Our study provides useful insight into appropriate experimental design and computational methods for serum biomarker discovery using LC-MS/MS based metabolomics. This study has led to the identification of candidate biomarkers with significant changes in metabolite levels between HCC cases and cirrhotic controls. This is the first MS-based metabolic biomarker discovery study on Egyptian subjects that led to the identification of candidate metabolites that discriminate early stage HCC from patients with liver cirrhosis.

Bayesian variable selection for the analysis of microarray data with censored outcomes

MOTIVATION A common task in microarray data analysis consists of identifying genes associated with a phenotype. When the outcomes of interest are censored time-to-event data, standard approaches assess the effect of genes by fitting univariate survival models. In this paper, we propose a Bayesian variable selection approach, which allows the identification of relevant markers by jointly assessing sets of genes. We consider accelerated failure time (AFT) models with log-normal and log-t distributional assumptions. A data augmentation approach is used to impute the failure times of censored observations and mixture priors are used for the regression coefficients to identify promising subsets of variables. The proposed method provides a unified procedure for the selection of relevant genes and the prediction of survivor functions. RESULTS We demonstrate the performance of the method on simulated examples and on several microarray datasets. For the simulation study, we consider scenarios with large number of noisy variables and different degrees of correlation between the relevant and non-relevant (noisy) variables. We are able to identify the correct covariates and obtain good prediction of the survivor functions. For the microarray applications, some of our selected genes are known to be related to the diseases under study and a few are in agreement with findings from other researchers. AVAILABILITY The Matlab code for implementing the Bayesian variable selection method may be obtained from the corresponding author. CONTACT mvannucci@stat.tamu.edu SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Atheroprotective lipoprotein effects of a niacin-simvastatin combination compared to low- and high-dose simvastatin monotherapy.

BACKGROUND Niacin has multiple lipoprotein effects that may provide cardiovascular benefit when added to statin monotherapy. METHODS In this randomized, placebo-controlled trial (n = 75) of magnetic resonance imaging of carotid atherosclerosis, we performed a secondary comparison of combination niacin-statin (simvastatin 20 mg/Niacin-ER 2G [S20/N]) to monotherapy with moderate (20 mg [S20]) and high-dose (80 mg [S80]) simvastatin on lipids, apolipoproteins (apo), low density lipoprotein (LDL) and high density lipoprotein (HDL) particle subclasses, and inflammatory markers. RESULTS At baseline, average age was 71, 72% were male, 62.5% used statins, and average LDL-cholesterol was 111 mg/dL. At 12 months, S20/N, compared to S80, significantly reduced apoB (-36.6% vs -11.9%; P = .05) and lipoprotein(a) (-18% vs +3.5%; P = .001) and had at least an equivalent effect on LDL-cholesterol (-39.3% vs -24.3%; P = .24). The combination reduced the proportion of subjects with atherogenic LDL pattern-B (50% to 11.5%) compared to S80 (56% to 56%) (P = .01). Despite increases in plasma free fatty acids (+62.4%; F = 5.65, P = .005 vs S20 and S80), plasma triglycerides (-29.4%; F = 6.88, P = .002 vs S20 and S80), and very-low-density lipoprotein (-44.2%; F = 7.94, P < .001 vs S20 and S80), levels were reduced by S20/N. S20/N increased HDL-cholesterol levels (+18.1%) as compared to S20 (0%) and S80 (+5.9%) (P < .001 vs both statin arms), largely due to an increase in HDL particle size (+4.6%; P = .01 vs both statin arms). CONCLUSIONS We demonstrate that full-dose niacin/moderate-dose simvastatin combination has sustained benefits on atherogenic apoB lipoproteins, at least comparable to high-dose simvastatin, while also raising HDL-cholesterol. Results of large clinical trials will inform whether niacin-statin combinations reduce cardiovascular disease events.

The Epidemiology of Sleep Quality, Sleep Patterns, Consumption of Caffeinated Beverages, and Khat Use among Ethiopian College Students

Objective. To evaluate sleep habits, sleep patterns, and sleep quality among Ethiopian college students; and to examine associations of poor sleep quality with consumption of caffeinated beverages and other stimulants. Methods. A total of 2,230 undergraduate students completed a self-administered comprehensive questionnaire which gathered information about sleep complaints, sociodemographic and lifestyle characteristics,and theuse of caffeinated beverages and khat. We used multivariable logistic regression procedures to estimate odds ratios for the associations of poor sleep quality with sociodemographic and behavioral factors. Results. Overall 52.7% of students were classified as having poor sleep quality (51.8% among males and 56.9% among females). In adjusted multivariate analyses, caffeine consumption (OR = 1.55; 95% CI: 1.25–1.92), cigarette smoking (OR = 1.68; 95% CI: 1.06–2.63), and khat use (OR = 1.72, 95% CI: 1.09–2.71) were all associated with increased odds of long-sleep latency (>30 minutes). Cigarette smoking (OR = 1.74; 95% CI: 1.11–2.73) and khat consumption (OR = 1.91; 95% CI: 1.22–3.00) were also significantly associated with poor sleep efficiency (<85%), as well as with increased use of sleep medicine. Conclusion. Findings from the present study demonstrate the high prevalence of poor sleep quality and its association with stimulant use among college students. Preventive and educational programs for students should include modules that emphasize the importance of sleep and associated risk factors.

Resistin gene variation is associated with systemic inflammation but not plasma adipokine levels, metabolic syndrome or coronary atherosclerosis in nondiabetic Caucasians.

Objective  Resistin causes insulin resistance and diabetes in mice whereas in humans it is linked to inflammation and atherosclerosis. Few human genetic studies of resistin in inflammation and atherosclerosis have been performed. We hypothesized that the –420C>G putative gain‐of‐function resistin variant would be associated with inflammatory markers and atherosclerosis but not with metabolic syndrome or adipokines in humans.