Integrating metabolomics profiling measurements across multiple biobanks.

To optimize the quality of large scale mass-spectrometry based metabolomics data obtained from semiquantitative profiling measurements, it is important to use a strategy in which dedicated measurement designs are combined with a strict statistical quality control regime. This assures consistently high-quality results across measurements from individual studies, but semiquantitative data have been so far only comparable for samples measured within the same study. To enable comparability and integration of semiquantitative profiling data from different large scale studies over the time course of years, the measurement and quality control strategy has to be extended. We introduce a strategy to allow the integration of semiquantitative profiling data from different studies. We demonstrate that lipidomics data generated in samples from three different large biobanks acquired in the time course of 3 years can be effectively combined when using an appropriate measurement design and transfer model. This strategy paves the way toward an integrative usage of semiquantitative metabolomics data sets of multiple studies to validate biological findings in another study and/or to increase the statistical power for discovery of biomarkers or pathways by combining studies.

Discovery of early-stage biomarkers for diabetic kidney disease using ms-based metabolomics (FinnDiane study).

Diabetic kidney disease (DKD) is a devastating complication that affects an estimated third of patients with type 1 diabetes mellitus (DM). There is no cure once the disease is diagnosed, but early treatment at a sub-clinical stage can prevent or at least halt the progression. DKD is clinically diagnosed as abnormally high urinary albumin excretion rate (AER). We hypothesize that subtle changes in the urine metabolome precede the clinically significant rise in AER. To test this, 52 type 1 diabetic patients were recruited by the FinnDiane study that had normal AER (normoalbuminuric). After an average of 5.5 years of follow-up half of the subjects (26) progressed from normal AER to microalbuminuria or DKD (macroalbuminuria), the other half remained normoalbuminuric. The objective of this study is to discover urinary biomarkers that differentiate the progressive form of albuminuria from non-progressive form of albuminuria in humans. Metabolite profiles of baseline 24 h urine samples were obtained by gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) to detect potential early indicators of pathological changes. Multivariate logistic regression modeling of the metabolomics data resulted in a profile of metabolites that separated those patients that progressed from normoalbuminuric AER to microalbuminuric AER from those patients that maintained normoalbuminuric AER with an accuracy of 75% and a precision of 73%. As this data and samples are from an actual patient population and as such, gathered within a less controlled environment it is striking to see that within this profile a number of metabolites (identified as early indicators) have been associated with DKD already in literature, but also that new candidate biomarkers were found. The discriminating metabolites included acyl-carnitines, acyl-glycines and metabolites related to tryptophan metabolism. We found candidate biomarkers that were univariately significant different. This study demonstrates the potential of multivariate data analysis and metabolomics in the field of diabetic complications, and suggests several metabolic pathways relevant for further biological studies. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s11306-011-0291-6) contains supplementary material, which is available to authorized users.

Analysis of human serum by liquid chromatography-mass spectrometry: improved sample preparation and data analysis.

Discovery of biomarkers is a fast developing field in proteomics research. Liquid chromatography coupled on line to mass spectrometry (LC-MS) has become a powerful method for the sensitive detection, quantification and identification of proteins and peptides in biological fluids like serum. However, the presence of highly abundant proteins often masks those of lower abundance and thus generally prevents their detection and identification in proteomics studies. To perform future comparative analyses of samples from a serum bank of cervical cancer patients in a longitudinal and cross-sectional manner, methodology based on the depletion of high-abundance proteins followed by tryptic digestion and LC-MS has been developed. Two sample preparation methods were tested in terms of their efficiency to deplete high-abundance serum proteins and how they affect the repeatability of the LC-MS data sets. The first method comprised depletion of human serum albumin (HSA) on a dye ligand chromatographic and immunoglobulin G (IgG) on an immobilized Protein A support followed by tryptic digestion, fractionation by cation-exchange chromatography, trapping on a C18 column and reversed-phase LC-MS. The second method included depletion of the six most abundant serum proteins based on multiple immunoaffinity chromatography followed by tryptic digestion, trapping on a C18 column and reversed-phase LC-MS. Repeatability of the overall procedures was evaluated in terms of retention time and peak area for a selected number of endogenous peptides showing that the second method, besides being less time consuming, gave more repeatable results (retention time: <0.1% RSD; peak area: <30% RSD). Application of an LC-MS component detection algorithm followed by principal component analysis (PCA) enabled discrimination of serum samples that were spiked with horse heart cytochrome C from non-spiked serum and the detection of a concentration trend, which correlated to the amount of spiked horse heart cytochrome C to a level of 5 pmol cytochrome C in 2 microl original serum.

Integrated statistical analysis of cDNA microarray and NIR spectroscopic data applied to a hemp dataset.

Both cDNA microarray and spectroscopic data provide indirect information about the chemical compounds present in the biological tissue under consideration. In this paper simple univariate and bivariate measures are used to investigate correlations between both types of high dimensional analyses. A large dataset of 42 hemp samples on which 3456 cDNA clones and 351 NIR wavelengths have been measured, was analyzed using graphical representations. For this purpose we propose clustered correlation and clustered discrimination images. Large, tissue-related differences are seen to dominate the cDNA-NIR correlation structure but smaller, more difficult to detect, variety-related differences can be found at specific cDNA clone/NIR wavelength combinations.

The influence of different structure representations on the clustering of an RNA nucleotides data set.

The last couple of years an overwhelming amount of data has emerged in the field of biomolecular structure determination. To explore information hidden in these structure databases, clustering techniques can be used. The outcome of the clustering experiments largely depends, among others, on the way the data is represented; therefore, the choice how to represent the molecular structure information is extremely important. This article describes what the influence of the different representations on the clustering is and how it can be analyzed by means of a dendrogram comparison method. All experiments are performed using a data set consisting of RNA trinucleotides. Besides the most basic structure representation, the Cartesian coordinates representation, several other structure representations are used.

Using genetic algorithms for the construction of phylogenetic trees: application to G-protein coupled receptor sequences.

Many different phylogenetic clustering techniques are used currently. One approach is to first determine the topology with a common clustering method and then calculate the branch lengths of the tree. If the resulting tree is not optimal exchanging tree branches can make some local changes in the tree topology. The whole process can be iterated until a satisfactory result has been obtained. The efficiency of this method fully depends on the initially generated tree. Although local changes are made, the optimal tree will never be found if the initial tree is poorly chosen. In this article, genetic algorithms are applied such that the optimal tree can be found even with a bad initial tree topology. This tree generating method is tested by comparing its results with the results of the FITCH program in the PHYLIP software package. Two simulated data sets and a real data set are used.