Erik Alm

The correspondence problem for metabonomics datasets.

In metabonomics it is difficult to tell which peak is which in datasets with many samples. This is known as the correspondence problem. Data from different samples are not synchronised, i.e., the peak from one metabolite does not appear in exactly the same place in all samples. For datasets with many samples, this problem is nontrivial, because each sample contains hundreds to thousands of peaks that shift and are identified ambiguously. Statistical analysis of the data assumes that peaks from one metabolite are found in one column of a data table. For every error in the data table, the statistical analysis loses power and the risk of missing a biomarker increases. It is therefore important to solve the correspondence problem by synchronising samples and there is no method that solves it once and for all. In this review, we analyse the correspondence problem, discuss current state-of-the-art methods for synchronising samples, and predict the properties of future methods.

A note on normalization of biofluid 1D 1H-NMR data

One-dimensional 1H nuclear magnetic resonance (1D 1H-NMR) has been used extensively as a metabolic profiling tool for investigating urine and other biological fluids. Under ideal conditions, 1H-NMR peak intensities are directly proportional to metabolite concentrations and thus are useful for class prediction and biomarker discovery. However, many biological, experimental and instrumental variables can affect absolute NMR peak intensities. Normalizing or scaling data to minimize the influence of these variables is a critical step in producing robust, reproducible analyses. Traditionally, analyses of biological fluids have relied on the total spectral area [constant sum (CS)] to normalize individual intensities. This approach can introduce considerable inter-sample variance as changes in any individual metabolite will affect the scaling of all of the observed intensities. To reduce normalization-related variance, we have developed a histogram matching (HM) approach adapted from the field of image processing. We validate our approach using mixtures of synthetic compounds that mimic a biological extract and apply the method to an analysis of urine from rats treated with ethionine. We show that HM is a robust method for normalizing 1H-NMR data and propose it as an alternative to the traditional CS method.

A solution to the 1D NMR alignment problem using an extended generalized fuzzy Hough transform and mode support

AbstractThis paper approaches the problem of intersample peak correspondence in the context of later applying statistical data analysis techniques to 1D 1H-nuclear magnetic resonance (NMR) data. Any data analysis methodology will fail to produce meaningful results if the analyzed data table is not synchronized, i.e., each analyzed variable frequency (Hz) does not originate from the same chemical source throughout the entire dataset. This is typically the case when dealing with NMR data from biological samples. In this paper, we present a new state of the art for solving this problem using the generalized fuzzy Hough transform (GFHT). This paper describes significant improvements since the method was introduced for NMR datasets of plasma in Csenki et al. (Anal Bioanal Chem 389:875-885, 15) and is now capable of synchronizing peaks from more complex datasets such as urine as well as plasma data. We present a novel way of globally modeling peak shifts using principal component analysis, a new algorithm for calculating the transform and an effective peak detection algorithm. The algorithm is applied to two real metabonomic 1H-NMR datasets and the properties of the method are compared to bucketing. We implicitly prove that GFHT establishes the objectively true correspondence. Desirable features of the GFHT are: (1) intersample peak correspondence even if peaks change order on the frequency axis and (2) the method is symmetric with respect to the samples. FigureFrom chaos to order: heatmaps of a H-NMR spectral segment prior and post sorting on one peak position. Post sorting sample order reveals that peak positions exhibits distinctive patterns which are modeled by the GFHT to establish correspondence.

Automated annotation and quantification of metabolites in (1)H NMR data of biological origin

In 1H NMR metabolomic datasets, there are often over a thousand peaks per spectrum, many of which change position drastically between samples. Automatic alignment, annotation, and quantification of all the metabolites of interest in such datasets have not been feasible. In this work we propose a fully automated annotation and quantification procedure which requires annotation of metabolites only in a single spectrum. The reference database built from that single spectrum can be used for any number of 1H NMR datasets with a similar matrix. The procedure is based on the generalized fuzzy Hough transform (GFHT) for alignment and on Principal-components analysis (PCA) for peak selection and quantification. We show that we can establish quantities of 21 metabolites in several 1H NMR datasets and that the procedure is extendable to include any number of metabolites that can be identified in a single spectrum. The procedure speeds up the quantification of previously known metabolites and also returns a table containing the intensities and locations of all the peaks that were found and aligned but not assigned to a known metabolite. This enables both biopattern analysis of known metabolites and data mining for new potential biomarkers among the unknowns.

Time-resolved biomarker discovery in 1H-NMR data using generalized fuzzy Hough transform alignment and parallel factor analysis

This work addresses the subject of time-series analysis of comprehensive 1H-NMR data of biological origin. One of the problems with toxicological and efficacy studies is the confounding of correlation between the administered drug, its metabolites and the systemic changes in molecular dynamics, i.e., the flux of drug-related molecules correlates with the molecules of system regulation. This correlation poses a problem for biomarker mining since this confounding must be untangled in order to separate true biomarker molecules from dose-related molecules. One way of achieving this goal is to perform pharmacokinetic analysis. The difference in pharmacokinetic time profiles of different molecules can aid in the elucidation of the origin of the dynamics, this can even be achieved regardless of whether the identity of the molecule is known or not. This mode of analysis is the basis for metabonomic studies of toxicology and efficacy. One major problem concerning the analysis of 1H-NMR data generated from metabonomic studies is that of the peak positional variation and of peak overlap. These phenomena induce variance in the data, obscuring the true information content and are hence unwanted but hard to avoid. Here, we show that by using the generalized fuzzy Hough transform spectral alignment, variable selection, and parallel factor analysis, we can solve both the alignment and the confounding problem stated above. Using the outlined method, several different temporal concentration profiles can be resolved and the majority of the studied molecules and their respective fluxes can be attributed to these resolved kinetic profiles. The resolved time profiles hereby simplifies finding true biomarkers and bio-patterns for early detection of biological conditions as well as providing more detailed information about the studied biological system. The presented method represents a significant step forward in time-series analysis of biological 1H-NMR data as it provides almost full automation of the whole data analysis process and is able to analyze over 800 unique features per sample. The method is demonstrated using a 1H-NMR rat urine dataset from a toxicology study and is compared with a classical approach: COW alignment followed by bucketing.