Helena U. Zacharias

Disease burden and risk profile in referred patients with moderate chronic kidney disease: composition of the German Chronic Kidney Disease (GCKD) cohort

BACKGROUND A main challenge for targeting chronic kidney disease (CKD) is the heterogeneity of its causes, co-morbidities and outcomes. Patients under nephrological care represent an important reference population, but knowledge about their characteristics is limited. METHODS We enrolled 5217 carefully phenotyped patients with moderate CKD [estimated glomerular filtration rate (eGFR) 30-60 mL/min per 1.73 m(2) or overt proteinuria at higher eGFR] under routine care of nephrologists into the German Chronic Kidney Disease (GCKD) study, thereby establishing the currently worldwide largest CKD cohort. RESULTS The cohort has 60% men, a mean age (±SD) of 60 ± 12 years, a mean eGFR of 47 ± 17 mL/min per 1.73 m(2) and a median (IQR) urinary albumin/creatinine ratio of 51 (9-392) mg/g. Assessment of causes of CKD revealed a high degree of uncertainty, with the leading cause unknown in 20% and frequent suspicion of multifactorial pathogenesis. Thirty-five per cent of patients had diabetes, but only 15% were considered to have diabetic nephropathy. Cardiovascular disease prevalence was high (32%, excluding hypertension); prevalent risk factors included smoking (59% current or former smokers) and obesity (43% with BMI >30). Despite widespread use of anti-hypertensive medication, only 52% of the cohort had an office blood pressure <140/90 mmHg. Family histories for cardiovascular events (39%) and renal disease (28%) suggest familial aggregation. CONCLUSIONS Patients with moderate CKD under specialist care have a high disease burden. Improved diagnostic accuracy, rigorous management of risk factors and unravelling of the genetic predisposition may represent strategies for improving prognosis.

Performance Evaluation of Algorithms for the Classification of Metabolic 1H NMR Fingerprints

Nontargeted metabolite fingerprinting is increasingly applied to biomedical classification. The choice of classification algorithm may have a considerable impact on outcome. In this study, employing nested cross-validation for assessing predictive performance, six binary classification algorithms in combination with different strategies for data-driven feature selection were systematically compared on five data sets of urine, serum, plasma, and milk one-dimensional fingerprints obtained by proton nuclear magnetic resonance (NMR) spectroscopy. Support Vector Machines and Random Forests combined with t-score-based feature filtering performed well on most data sets, whereas the performance of the other tested methods varied between data sets.

Data Normalization of 1H NMR Metabolite Fingerprinting Data Sets in the Presence of Unbalanced Metabolite Regulation

Data normalization is an essential step in NMR-based metabolomics. Conducted properly, it improves data quality and removes unwanted biases. The choice of the appropriate normalization method is critical and depends on the inherent properties of the data set in question. In particular, the presence of unbalanced metabolic regulation, where the different specimens and cohorts under investigation do not contain approximately equal shares of up- and down-regulated features, may strongly influence data normalization. Here, we demonstrate the suitability of the Shapiro-Wilk test to detect such unbalanced regulation. Next, employing a Latin-square design consisting of eight metabolites spiked into a urine specimen at eight different known concentrations, we show that commonly used normalization and scaling methods fail to retrieve true metabolite concentrations in the presence of increasing amounts of glucose added to simulate unbalanced regulation. However, by learning the normalization parameters on a subset of nonregulated features only, Linear Baseline Normalization, Probabilistic Quotient Normalization, and Variance Stabilization Normalization were found to account well for different dilutions of the samples without distorting the true spike-in levels even in the presence of marked unbalanced metabolic regulation. Finally, the methods described were applied successfully to a real world example of unbalanced regulation, namely, a set of plasma specimens collected from patients with and without acute kidney injury after cardiac surgery with cardiopulmonary bypass use.

Identification of Plasma Metabolites Prognostic of Acute Kidney Injury after Cardiac Surgery with Cardiopulmonary Bypass

Acute kidney injury (AKI) is a frequent complication after cardiopulmonary bypass, but early detection of postoperative AKI remains challenging. Protein biomarkers predict AKI excellently in homogeneous cohorts but are less reliable in patients suffering from various comorbidities. We employed nuclear magnetic resonance spectroscopy in a prospective study of 85 adult cardiac surgery patients to identify metabolites prognostic of AKI in plasma specimens collected 24 h after surgery. Postoperative AKI of stages 1-3, as defined by the Acute Kidney Injury Network (AKIN), developed in 33 cases. A random forests classifier trained on the NMR spectra prognosticated AKI across all stages, with an average accuracy of 80 ± 0.9% and an area under the receiver operating characteristic curve of 0.87 ± 0.01. Prognostications were based, on average, on 24 ± 2.8 spectral features. Among the set of discriminative ions and molecules identified were Mg(2+), lactate, and the glucuronide conjugate of propofol. Using creatinine, Mg(2+), and lactate levels to derive an AKIN index score, we found AKIN 1 disease to be largely indistinguishable from AKIN 0, in concordance with the rather mild nature of AKIN 1 disease.

Current Experimental, Bioinformatic and Statistical Methods used in NMR Based Metabolomics

The aim of this contribution is to familiarize the reader with experimental, bioinformatic and statistical strategies currently used in the field of solution NMR based metabolomics. Special emphasis is given to methods that have worked well in our hands. Methods covered include sample preparation, acquisition and processing of NMR spectra, and identification and quantification of metabolites. Further consideration is given to data normalization and scaling, unsupervised and supervised statistical data analysis, the biomedical interpretation of results, and the centralized community-wide storage and retrieval of NMR data.

Visceral adipose tissue but not subcutaneous adipose tissue is associated with urine and serum metabolites

Obesity is a complex multifactorial phenotype that influences several metabolic pathways. Yet, few studies have examined the relations of different body fat compartments to urinary and serum metabolites. Anthropometric phenotypes (visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT), the ratio between VAT and SAT (VSR), body mass index (BMI), waist circumference (WC)) and urinary and serum metabolite concentrations measured by nuclear magnetic resonance spectroscopy were measured in a population-based sample of 228 healthy adults. Multivariable linear and logistic regression models, corrected for multiple testing using the false discovery rate, were used to associate anthropometric phenotypes with metabolites. We adjusted for potential confounding variables: age, sex, smoking, physical activity, menopausal status, estimated glomerular filtration rate (eGFR), urinary glucose, and fasting status. In a fully adjusted logistic regression model dichotomized for the absence or presence of quantifiable metabolite amounts, VAT, BMI and WC were inversely related to urinary choline (ß = -0.18, p = 2.73*10−3), glycolic acid (ß = -0.20, 0.02), and guanidinoacetic acid (ß = -0.12, p = 0.04), and positively related to ethanolamine (ß = 0.18, p = 0.02) and dimethylamine (ß = 0.32, p = 0.02). BMI and WC were additionally inversely related to urinary glutamine and lactic acid. Moreover, WC was inversely associated with the detection of serine. VAT, but none of the other anthropometric parameters, was related to serum essential amino acids, such as valine, isoleucine, and phenylalanine among men. Compared to other adiposity measures, VAT demonstrated the strongest and most significant relations to urinary and serum metabolites. The distinct relations of VAT, SAT, VSR, BMI, and WC to metabolites emphasize the importance of accurately differentiating between body fat compartments when evaluating the potential role of metabolic regulation in the development of obesity-related diseases, such as insulin resistance, type 2 diabetes, and cardiovascular disease.

Evaluation of dilution and normalization strategies to correct for urinary output in HPLC-HRTOFMS metabolomics

AbstractReliable identification of features distinguishing biological groups of interest in urinary metabolite fingerprints requires the control of total metabolite abundance, which may vary significantly as the kidneys adjust the excretion of water and solutes to meet the homeostatic needs of the body. Failure to account for such variation may lead to misclassification and accumulation of missing data in case of less concentrated urine specimens. Here, different pre- and post-acquisition methods of normalization were compared systematically for their ability to recover features from liquid chromatography-mass spectrometry metabolite fingerprints of urine that allow distinction between patients with chronic kidney disease and healthy controls. Methods of normalization that were employed prior to analysis included dilution of urine specimens to either a fixed creatinine concentration or osmolality value. Post-acquisition normalization methods applied to chromatograms of 1:4 diluted urine specimens comprised normalization to creatinine, osmolality, and sum of all integrals. Dilution of urine specimens to a fixed creatinine concentration resulted not only in the least number of missing values, but it was also the only method allowing the unambiguous classification of urine specimens from healthy and diseased individuals. The robustness of classification could be confirmed for two independent patient cohorts of chronic kidney disease patients and yielded a shared set of 49 discriminant metabolite features. Graphical AbstractDilution to a uniform creatinine concentration across urine specimens yields more comparable urinary metabolite fingerprints

Quantification of Metabolites by NMR Spectroscopy in the Presence of Protein

The high reliability of NMR spectroscopy makes it an ideal tool for large-scale metabolomic studies. However, the complexity of biofluids and, in particular, the presence of macromolecules poses a significant challenge. Ultrafiltration and protein precipitation are established means of deproteinization and recovery of free or total metabolite content, but neither is ever complete. In addition, aside from cost and labor, all deproteinization methods constitute an additional source of experimental variation. The Carr-Purcell-Meiboom-Gill (CPMG) echo-train acquisition of NMR spectra obviates the need for prior deproteinization by attenuating signals from macromolecules, but concentration values of metabolites measured in blood plasma will not necessarily reflect total or free metabolite content. Here, in contrast to approaches that propose the determination of individual T1 and T2 relaxation times for the computation of correction factors, we demonstrate their determination by spike-in experiments with known amounts of metabolites in pooled samples of the matrix of interest to facilitate the measurement of total metabolite content. Provided that the protein content does not vary too much among individual samples, accurate quantitation of metabolites is feasible. Moreover, samples with significantly deviating protein content may be readily recognized by inclusion of a standard that shows moderate protein binding. It is also shown that urinary proteins when present in high concentrations may effect detection of common urinary metabolites prone to strong protein binding such as tryptophan.

Scale-Invariant Biomarker Discovery in Urine and Plasma Metabolite Fingerprints

Metabolomics data is typically scaled to a common reference like a constant volume of body fluid, a constant creatinine level, or a constant area under the spectrum. Such scaling of the data, however, may affect the selection of biomarkers and the biological interpretation of results in unforeseen ways. Here, we studied how both the outcome of hypothesis tests for differential metabolite concentration and the screening for multivariate metabolite signatures are affected by the choice of scale. To overcome this problem for metabolite signatures and to establish a scale-invariant biomarker discovery algorithm, we extended linear zero-sum regression to the logistic regression framework and showed in two applications to 1H NMR-based metabolomics data how this approach overcomes the scaling problem. Logistic zero-sum regression is available as an R package as well as a high-performance computing implementation that can be downloaded at https://github.com/rehbergT/zeroSum .

Dilution correction for dynamically influenced urinary analyte data

Urinary analyte data has to be corrected for the sample specific dilution as the dilution varies intra- and interpersonally dramatically, leading to non-comparable concentration measures. Most methods of dilution correction utilized nowadays like probabilistic quotient normalization or total spectra normalization result in a division of the raw data by a dilution correction factor. Here, however, we show that the implicit assumption behind the application of division, log-linearity between the urinary flow rate and the raw urinary concentration, does not hold for analytes which are not in steady state in blood. We explicate the physiological reason for this short-coming in mathematical terms and demonstrate the empirical consequences via simulations and on multiple time-point metabolomic data, showing the insufficiency of division-based normalization procedures to account for the complex non-linear analyte specific dependencies on the urinary flow rate. By reformulating normalization as a regression problem, we propose an analyte specific way to remove the dilution variance via a flexible non-linear regression methodology which then was shown to be more effective in comparison to division-based normalization procedures. In the progress, we developed several, easily applicable methods of normalization diagnostics to decide on the method of dilution correction in a given sample. On the way, we identified furthermore the time-span since last urination as an important variance factor in urinary metabolome data which is until now completely neglected. In conclusion, we present strong theoretical and empirical evidence that normalization has to be analyte specific in dynamically influenced data. Accordingly, we developed a normalization methodology for removing the dilution variance in urinary data respecting the single analyte kinetics.