Statistical search space reduction and two-dimensional data display approaches for UPLC-MS in biomarker discovery and pathway analysis

A new analytical strategy for biomarker recovery from directly coupled ultra-performance liquid chromatography time-of-flight mass spectrometry (UPLC Tof MS) data on biofluids is presented and exemplified using a study on hydrazine-induced liver toxicity. A key step in the strategy involves a novel procedure for reducing the spectroscopic search space by differential analysis of cohorts of normal and pathological samples using an orthogonal projection to latent structures discriminant analysis (O-PLS-DA). This efficiently sorts principal discriminators of toxicity from the background of thousands of metabolic features commonly observed in data sets generated by UPLC-MS analysis of biological fluids and is thus a powerful tool for biomarker discovery.     

Statistical total correlation spectroscopy scaling for enhancement of metabolic information recovery in biological NMR spectra

The high level of complexity in nuclear magnetic resonance (NMR) metabolic spectroscopic data sets has fueled the development of experimental and mathematical techniques that enhance latent biomarker recovery and improve model interpretability. We previously showed that statistical total correlation spectroscopy (STOCSY) can be used to edit NMR spectra to remove drug metabolite signatures that obscure metabolic variation of diagnostic interest. Here, we extend this "STOCSY editing" concept to a generalized scaling procedure for NMR data that enhances recovery of latent biochemical information and improves biological classification and interpretation. We call this new procedure STOCSY-scaling (STOCSY(S)). STOCSY(S) exploits the fixed proportionality in a set of NMR spectra between resonances from the same molecule to suppress or enhance features correlated with a resonance of interest. We demonstrate this new approach using two exemplar data sets: (a) a streptozotocin rat model (n = 30) of type 1 diabetes and (b) a human epidemiological study utilizing plasma NMR spectra of patients with metabolic syndrome (n = 67). In both cases significant biomarker discovery improvement was observed by using STOCSY(S): the approach successfully suppressed interfering NMR signals from glucose and lactate that otherwise dominate the variation in the streptozotocin study, which then allowed recovery of biomarkers such as glycine, which were otherwise obscured. In the metabolic syndrome study, we used STOCSY(S) to enhance variation from the high-density lipoprotein cholesterol peak, improving the prediction of individuals with metabolic syndrome from controls in orthogonal projections to latent structures discriminant analysis models and facilitating the biological interpretation of the results. Thus, STOCSY(S) is a versatile technique that is applicable in any situation in which variation, either biological or otherwise, dominates a data set at the expense of more interesting or important features. This approach is generally appropriate for many types of NMR-based complex mixture analyses and hence for wider applications in bioanalytical science.     

A strategy for identifying differences in large series of metabolomic samples analyzed by GC/MS

In metabolomics, the purpose is to identify and quantify all the metabolites in a biological system. Combined gas chromatography and mass spectrometry (GC/MS) is one of the most commonly used techniques in metabolomics together with 1H NMR, and it has been shown that more than 300 compounds can be distinguished with GC/MS after deconvolution of overlapping peaks. To avoid having to deconvolute all analyzed samples prior to multivariate analysis of the data, we have developed a strategy for rapid comparison of nonprocessed MS data files. The method includes baseline correction, alignment, time window determinations, alternating regression, PLS-DA, and identification of retention time windows in the chromatograms that explain the differences between the samples. Use of alternating regression also gives interpretable loadings, which retain the information provided by m/z values that vary between the samples in each retention time window. The method has been applied to plant extracts derived from leaves of different developmental stages and plants subjected to small changes in day length. The data show that the new method can detect differences between the samples and that it gives results comparable to those obtained when deconvolution is applied prior to the multivariate analysis. We suggest that this method can be used for rapid comparison of large sets of GC/MS data, thereby applying time-consuming deconvolution only to parts of the chromatograms that contribute to explain the differences between the samples.     

Monitoring diet effects via biofluids and their implications for metabolomics studies

The effect of diet on metabolites found in rat urine samples has been investigated using nuclear magnetic resonance (NMR) and a new ambient ionization mass spectrometry experiment, extractive electrospray ionization mass spectrometry (EESI-MS). Urine samples from rats with three different dietary regimens were readily distinguished using multivariate statistical analysis on metabolites detected by NMR and MS. To observe the effect of diet on metabolic pathways, metabolites related to specific pathways were also investigated using multivariate statistical analysis. Discrimination is increased by making observations on restricted compound sets. Changes in diet at 24-h intervals led to predictable changes in the spectral data. Principal component analysis was used to separate the rats into groups according to their different dietary regimens using the full NMR, EESI-MS data or restricted sets of peaks in the mass spectra corresponding only to metabolites found in the urea cycle and metabolism of amino groups pathway. By contrast, multivariate analysis of variance from the score plots showed that metabolites of purine metabolism obscure the classification relative to the full metabolite set. These results suggest that it may be possible to reduce the number of statistical variables used by monitoring the biochemical variability of particular pathways. It should also be possible by this procedure to reduce the effect of diet in the biofluid samples for such purposes as disease detection.     

Heteronuclear 1H-31P statistical total correlation NMR spectroscopy of intact liver for metabolic biomarker assignment: application to galactosamine-induced hepatotoxicity

As part of our ongoing development of methods for enhanced biomarker information recovery from spectroscopic data we present the first example of a new hetero-nuclear statistical total correlation spectroscopy (HET-STOCSY) approach applied to intact tissue samples collected as part of a toxicological study. One-dimensional 1H and 31P-{1H} magic angle spinning (MAS) NMR spectra of intact liver samples after galactosamine (galN) treatment to rats and after cotreatment of galN plus uridine were collected at 275 K. Individual samples were also followed by 1H and 31P-{1H} MAS NMR through time generating time dependent modulations in metabolite signatures relating to toxicity. High-resolution 1H NMR spectra of urine and plasma and clinical chemical data were also collected to establish a biological framework in which to place these novel statistical heterospectroscopic data. In HET-STOCSY, calculation of the covariance between the 31P-{1H} and 1H NMR signals of phosphorus containing metabolites allows their molecular connectivities to be established and the construction of virtual two-dimensional heteronuclear correlation spectra that connect all protons on the molecule to the heteroatom. We show how HET-STOCSY applied to MAS NMR spectra of liver samples can be used to augment biomarker detection. This approach is generic and can be applied to correlate the covarying signals for any spin-active nuclei where there is parallel or serial collection of data.     

Assessment of compatibility between extraction methods for NMR- and LC/MS-based metabolomics

Because of the wide range of chemically and structurally diverse metabolites, efforts to survey the complete metabolome rely on the implementation of multiplatform approaches based on nuclear magnetic resonance (NMR) and mass spectrometry (MS). Sample preparation disparities between NMR and MS, however, may limit the analysis of the same samples by both platforms. Specifically, deuterated solvents used in NMR strategies can complicate LC/MS analysis as a result of potential mass shifts, whereas acidic solutions typically used in LC/MS methods to enhance ionization of metabolites can severely affect reproducibility of NMR measurements. These intrinsically different sample preparation requirements result in the application of different procedures for metabolite extraction, which involve additional sample and unwanted variability. To address this issue, we investigated 12 extraction protocols in liver tissue involving different aqueous/organic solvents and temperatures that may satisfy the requirements for both NMR and LC/MS simultaneously. We found that deuterium exchange did not affect LC/MS results, enabling the measurement of metabolites by NMR and, subsequently, the direct analysis of the same samples by using LC/MS with no need for solvent exchange. Moreover, our results show that the choice of solvents rather than the temperature determined the extraction efficiencies of metabolites, a combination of methanol/chloroform/water and methanol/water being the extraction methods that best complement NMR and LC/MS analysis for metabolomic studies.     

Assessment of analytical reproducibility of 1H NMR spectroscopy based metabonomics for large-scale epidemiological research: the INTERMAP Study

Large-scale population phenotyping for molecular epidemiological studies is subject to all the usual criteria of analytical chemistry. As part of a major phenotyping investigation we have used high-resolution 1H NMR spectroscopy to characterize 24-h urine specimens obtained from population samples in Aito Town, Japan (n = 259), Chicago, IL (n = 315), and Guangxi, China (n = 278). We have investigated analytical reproducibility, urine specimen storage procedures, interinstrument variability, and split specimen detection. Our data show that the multivariate analytical reproducibility of the NMR screening platform was >98% and that most classification errors were due to urine specimen handling inhomogeneity. Differences in metabolite profiles were then assessed for Aito Town, Chicago, and Guangxi population samples; novel combinations of biomarkers were detected that separated the population samples. These cross-population differences in urinary metabolites could be related to genetic, dietary, and gut microbial factors.     

Advancing NMR sensitivity for LC-NMR-MS using a cryoflow probe: application to the analysis of acetaminophen metabolites in urine

Cryogenic cooling of the NMR radio frequency coils and electronics to give greatly enhanced sensitivity is arguably the most significant recent advance in NMR spectroscopy. Here we report the first cryogenic probe built in flow configuration and demonstrate the application to LC-NMR-MS studies. This probe provides superior sensitivity over conventional noncryogenic flow NMR probes, allowing the use of 100 microL of untreated urine (40% less material than previous studies that required preconcentration) and yet revealing drug metabolites hitherto undetected by LC-NMR-MS at 500 MHz. Besides the known sulfate and glucuronide metabolites, previously undetected metabolites of acetaminophen were directly observable in a 15-min on-flow experiment. Simultaneous MS data also provided knowledge on the NMR-silent functional moieties. Further, stop-flow LC-NMR-MS experiments were conducted for greater signal-to-noise ratios on minor metabolites. The cryoflow probe enables the NMR analysis of lower concentrations of metabolites than was previously possible for untreated biofluids. This strategy is generally applicable for samples containing mass-limited analytes, such as those from drug metabolism studies, biomarker and toxicity profiling, impurity analysis, and natural product analysis.     

Analysis of metabolomic data using support vector machines

Metabolomics is an emerging field providing insight into physiological processes. It is an effective tool to investigate disease diagnosis or conduct toxicological studies by observing changes in metabolite concentrations in various biofluids. Multivariate statistical analysis is generally employed with nuclear magnetic resonance (NMR) or mass spectrometry (MS) data to determine differences between groups (for instance diseased vs healthy). Characteristic predictive models may be built based on a set of training data, and these models are subsequently used to predict whether new test data falls under a specific class. In this study, metabolomic data is obtained by doing a (1)H NMR spectroscopy on urine samples obtained from healthy subjects (male and female) and patients suffering from Streptococcus pneumoniae. We compare the performance of traditional PLS-DA multivariate analysis to support vector machines (SVMs), a technique widely used in genome studies on two case studies: (1) a case where nearly complete distinction may be seen (healthy versus pneumonia) and (2) a case where distinction is more ambiguous (male versus female). We show that SVMs are superior to PLS-DA in both cases in terms of predictive accuracy with the least number of features. With fewer number of features, SVMs are able to give better predictive model when compared to that of PLS-DA.     

Combined top-down and bottom-up mass spectrometric approach to characterization of biomarkers for renal disease

Here we describe a mass spectrometry (MS) approach for biomarker discovery and structural characterization, based on both top-down and bottom-up analyses. Capillary electrophoresis (CE) coupled to electrospray ionization (ESI) time-of-flight (TOF) MS serves to separate and mass-measure the thousands of polypeptides contained in human urine. Statistical analysis of the differences between healthy control samples and patients with focal-segmental glomerulosclerosis, membranous glomerulonephritis, minimal change disease, IgA nephropathy, and diabetic nephropathy validates multiple biomarkers for the control and each of the diseases. To identify those biomarkers, we employ preparative CE, enabling direct infusion ESI MS analysis, followed by sample manipulation and reanalysis where necessary. We show how tandem Fourier transform ion cyclotron resonance (FT-ICR) MS identifies these sometimes large (>8 kDa) biomarkers. Critically, we maintain connectivity between the CE TOF MS data and the ICR data used for biomarker identification.     

Software tools for analysis of mass spectrometric lipidome data

New software tools for quantitative analysis of mass spectrometric lipidome data have been developed. The LIMSA tool finds and integrates peaks in a mass spectrum, matches the peaks with a user-supplied list of expected lipids, corrects for overlap in their isotopic patterns, and quantifies the identified lipid species according to internal standards. Three different algorithms for isotopic correction (deconvolution) were implemented and compared. LIMSA has a convenient user interface and can be applied on any type of MS spectrum. Typically, analysis of one spectrum takes only a few seconds. The SECD tool, designed for analysis of LC-MS data sets, provides an intuitive and informative display of MS chromatograms as two-dimensional "maps" for visual inspection of the data and allows the user to extract mass spectra, to be further analyzed with LIMSA, from arbitrary regions of these maps. More reliable analysis of complex lipidome data with improved signal-to-noise ratio is obtained when compared to standard time-range averaged spectra. The functionality of these tools is demonstrated by analysis of standard mixtures as well as complex biological samples. The tools described here make accurate, high-throughput analysis of extensive sample sets feasible and are made available to the scientific community free of charge.     

High-throughput data analysis for detecting and identifying differences between samples in GC/MS-based metabolomic analyses

In metabolomics, the objective is to identify differences in metabolite profiles between samples. A widely used tool in metabolomics investigations is gas chromatography-mass spectrometry (GC/MS). More than 400 compounds can be detected in a single analysis, if overlapping GC/MS peaks are deconvoluted. However, the deconvolution process is time-consuming and difficult to automate, and additional processing is needed in order to compare samples. Therefore, there is a need to improve and automate the data processing strategy for data generated in GC/MS-based metabolomics; if not, the processing step will be a major bottleneck for high-throughput analyses. Here we describe a new semiautomated strategy using a hierarchical multivariate curve resolution approach that processes all samples simultaneously. The presented strategy generates (after appropriate treatment, e.g., multivariate analysis) tables of all the detected metabolites that differ in relative concentrations between samples. The processing of 70 samples took similar time to that of the GC/TOFMS analyses of the samples. The strategy has been validated using two different sets of samples: a complex mixture of standard compounds and Arabidopsis samples.     

"Add to subtract": a simple method to remove complex background signals from the 1H nuclear magnetic resonance spectra of mixtures

Because of its highly reproducible and quantitative nature and minimal requirements for sample preparation or separation, (1)H nuclear magnetic resonance (NMR) spectroscopy is widely used for profiling small-molecule metabolites in biofluids. However (1)H NMR spectra contain many overlapped peaks. In particular, blood serum/plasma and diabetic urine samples contain high concentrations of glucose, which produce strong peaks between 3.2 ppm and 4.0 ppm. Signals from most metabolites in this region are overwhelmed by the glucose background signals and become invisible. We propose a simple "Add to Subtract" background subtraction method and show that it can reduce the glucose signals by 98% to allow retrieval of the hidden information. This procedure includes adding a small drop of concentrated glucose solution to the sample in the NMR tube, mixing, waiting for an equilibration time, and acquisition of a second spectrum. The glucose-free spectra are then generated by spectral subtraction using Bruker Topspin software. Subsequent multivariate statistical analysis can then be used to identify biomarker candidate signals for distinguishing different types of biological samples. The principle of this approach is generally applicable for all quantitative spectral data and should find utility in a variety of NMR-based mixture analyses as well as in metabolite profiling.     

Statistical correlation and projection methods for improved information recovery from diffusion-edited NMR spectra of biological samples

Although NMR spectroscopic techniques coupled with multivariate statistics can yield much useful information for classifying biological samples based on metabolic profiles, biomarker identification remains a time-consuming and complex procedure involving separation methods, two-dimensional NMR, and other spectroscopic tools. We present a new approach to aid complex biomixture analysis that combines diffusion ordered (DO) NMR spectroscopy with statistical total correlation spectroscopy (STOCSY) and demonstrate its application in the characterization of urinary biomarkers and enhanced information recovery from plasma NMR spectra. This method relies on calculation and display of the covariance of signal intensities from the various nuclei on the same molecule across a series of spectra collected under different pulsed field gradient conditions that differentially attenuate the signal intensities according to translational molecular diffusion rates. We term this statistical diffusion-ordered spectroscopy (S-DOSY). We also have developed a new visualization tool in which the apparent diffusion coefficients from DO spectra are projected onto a 1D NMR spectrum (diffusion-ordered projection spectroscopy, DOPY). Both methods either alone or in combination have the potential for general applications to any complex mixture analysis where the sample contains compounds with a range of diffusion coefficients.     

A statistical model for identifying proteins by tandem mass spectrometry

A statistical model is presented for computing probabilities that proteins are present in a sample on the basis of peptides assigned to tandem mass (MS/MS) spectra acquired from a proteolytic digest of the sample. Peptides that correspond to more than a single protein in the sequence database are apportioned among all corresponding proteins, and a minimal protein list sufficient to account for the observed peptide assignments is derived using the expectation-maximization algorithm. Using peptide assignments to spectra generated from a sample of 18 purified proteins, as well as complex H. influenzae and Halobacterium samples, the model is shown to produce probabilities that are accurate and have high power to discriminate correct from incorrect protein identifications. This method allows filtering of large-scale proteomics data sets with predictable sensitivity and false positive identification error rates. Fast, consistent, and transparent, it provides a standard for publishing large-scale protein identification data sets in the literature and for comparing the results obtained from different experiments.     

Metabolite projection analysis for fast identification of metabolites in metabonomics. Application in an amiodarone study

The assignment of significantly changed NMR signals, which were identified with the help of multivariate models, to individual metabolites in biofluids is a manual and tedious task requiring knowledge in chemometrics and NMR spectroscopy. Metabolite projection analysis, introduced in this work, allows automatic linking of multivariate models with metabolites by skipping the level of manual NMR signal identification. The method depends on the projection of sets of metabolite NMR spectra from a database into PCA or PLS models of NMR spectra of biofluid samples. Metabolites that are significantly changed can be identified graphically in metabolite projection plots or numerically as projected virtual concentration. The method is demonstrated together with a newly introduced algorithm for refined nonequidistant binning using a metabonomics study with amiodarone as administered drug. Amiodarone can induce phospholipidosis in the lung and liver, which is accompanied by associated organ toxicity in these organs. It is shown how metabolite projection analysis allows easy and fast tentative assignment of all structures of metabolites whose concentrations in the urine samples significantly changed upon dosage. These metabolites had also been identified previously by manually interpreting the multivariate models and spectra. Among these metabolites, phenylacetylglycine was also identified as being significantly increased. This metabolite has recently been proposed as urinary biomarker for phospholipidosis.     

Robust data processing and normalization strategy for MALDI mass spectrometric imaging

Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI) provides localized information about the molecular content of a tissue sample. To derive reliable conclusions from MSI data, it is necessary to implement appropriate processing steps in order to compare peak intensities across the different pixels comprising the image. Here, we review commonly used normalization methods, and propose a rational data processing strategy, for robust evaluation and modeling of MSI data. The approach includes newly developed heuristic methods for selecting biologically relevant peaks and pixels to reduce the size of a data set and remove the influence of the applied MALDI matrix. The methods are demonstrated on a MALDI MSI data set of a sagittal section of rat brain (4750 bins, m/z = 50-1000, 111 × 185 pixels) and the proposed preferred normalization method uses the median intensity of selected peaks, which were determined to be independent of the MALDI matrix. This was found to effectively compensate for a range of known limitations associated with the MALDI process and irregularities in MS image sampling routines. This new approach is relevant for processing of all MALDI MSI data sets, and thus likely to have impact in biomarker profiling, preclinical drug distribution studies, and studies addressing underlying molecular mechanisms of tissue pathology.     

Optimization of human plasma 1H NMR spectroscopic data processing for high-throughput metabolic phenotyping studies and detection of insulin resistance related to type 2 diabetes

Optimizing NMR experimental parameters for high-throughput metabolic phenotyping requires careful examination of the total biochemical information obtainable from (1)H NMR data, which includes concentration and molecular dynamics information. Here we have applied two different types of mathematical transformation (calculation of the first derivative of the NMR spectrum and Gaussian shaping of the free-induction decay) to attenuate broad spectral features from macromolecules and enhance the signals of small molecules. By application of chemometric methods such as principal component analysis (PCA), orthogonal projections to latent structures discriminant analysis (O-PLS-DA) and statistical spectroscopic tools such as statistical total correlation spectroscopy (STOCSY), we show that these methods successfully identify the same potential biomarkers as spin-echo (1)H NMR spectra in which broad lines are suppressed via T2 relaxation editing. Finally, we applied these methods for identification of the metabolic phenotype of patients with type 2 diabetes. This "virtual" relaxation-edited spectroscopy (RESY) approach can be particularly useful for high-throughput screening of complex mixtures such as human plasma and may be useful for extraction of latent biochemical information from legacy or archived NMR data sets for which only standard 1D data sets exist.     

Sample handling for mass spectrometric proteomic investigations of human sera

Proteomic investigations of sera are potentially of value for diagnosis, prognosis, choice of therapy, and disease activity assessment by virtue of discovering new biomarkers and biomarker patterns. Much debate focuses on the biological relevance and the need for identification of such biomarkers while less effort has been invested in devising standard procedures for sample preparation and storage in relation to model building based on complex sets of mass spectrometric (MS) data. Thus, development of standardized methods for collection and storage of patient samples together with standards for transportation and handling of samples are needed. This requires knowledge about how sample processing affects MS-based proteome analyses and thereby how nonbiological biased classification errors are avoided. In this study, we characterize the effects of sample handling, including clotting conditions, storage temperature, storage time, and freeze/thaw cycles, on MS-based proteomics of human serum by using principal components analysis, support vector machine learning, and clustering methods based on genetic algorithms as class modeling and prediction methods. Using spiking to artificially create differentiable sample groups, this integrated approach yields data that--even when working with sample groups that differ more than may be expected in biological studies--clearly demonstrate the need for comparable sampling conditions for samples used for modeling and for the samples that are going into the test set group. Also, the study emphasizes the difference between class prediction and class comparison studies as well as the advantages and disadvantages of different modeling methods.