May 1, 2006 (Vol. 26, No. 9)

Finding Small Biological Variations in Large Data Sample Sets at Earlier Stage

Recent advances in technology are bringing metabolic profiling to the forefront of drug discovery. Although a lot of existing technology has been adapted for this field, researchers are developing unique methods to provide more qualitative and quantitative information. In addition, more stringent demands by the FDA are pushing metabolic profiling to earlier stages.

Addressing Data Processing

One of the biggest challenges in metabolic profiling is data analysis and processing. &#8220Youre trying to find very small biological variations in very big data sample sets,&#8220 says Julia Wingate, Ph.D., senior product manager, LC-MS business group at Applied Biosystems ( The companys most recent software, MarkerView 1.1, offers three types of statistical analysis for mass spectrometry data: principal component, t-test, and discriminate analysis. The software also allows users to export data into other third-party statistics packages.

Dr. Wingate stresses that the software reflects areas where the company has expertise&#8212extracting real information from LC-MS chromatograms, picking out real peaks, aligning and normalizing them, and also visualizing data and developing ways to link statistical analysis back to the raw mass spectrometry data. A new feature of this version enables the user to view MS-MS information and creates a list to go back and re-acquire more data for more in-depth analysis of certain compounds.

&#8220So if you see a few compounds that look like potential biomarkers, you can easily go back and view the underlying MS information and the MS-MS information, if it was acquired, to start getting structural information. MarkerView doesnt compete with existing statistical software, but its complementary, depending on the type of analyses people are doing at various stages of processing their data,&#8220 explains Dr. Wingate.

Advances in Technology

The Metabolic Profiler combines NMR, MS, and LC into one platform for metabolic profiling. Developed by Bruker BioSpin (, the system provides complementary information when searching for a new drug metabolite or a low-level metabolite.

&#8220A significant amount of the work is done by NMR alone,&#8220 says Werner Maas, Ph.D., vp of R&D. He explains that a single sample can have an unknown amount of low-level metabolites. &#8220Once you get to lower levels, around the 10-microMolar level, there are many metabolites that result in signatures that all overlap and add up as the NMR spectrum. Unraveling this information is a bit tricky.&#8220

To do that, the signatures of individual metabolites need to be known. The company has made this information available through its recently launched NMR Metabolic Profiler Database. &#8220Since the NMR signature changes as a function of pH of the sample, we characterize those shifts as a function of the compound. Once we know the pH of the sample, we can go back into the database and say, we have a pH of 6.3, so this is the spectrum we have to use to unravel the complete spectrum,&#8220 Dr. Maas states.

The database contains endogenous metabolites, as well as metabolites from aspirin, Tylenol, and dietary components. Pharmaceutical customers can customize the database and add their own metabolites.

Waters ( also provides technology for metabolic profiling. Their UPLC (Ultra Performance Liquid Chromatograph) systems are designed to provide maximum sensitivity, resolution, and speed.

&#8220We recommend using our UPLC and then using a time-of-flight mass spectrometer to provide accurate mass information about the samples. Then the data goes into our informatics system where it is extracted from the raw chromatograms. Various found compounds are put into a format where we can perform statistical analysis and see which ones could potentially be biomarkers,&#8220 says John Shockcor, Ph.D., business development manager, metabolic profiling.

He stresses that the system does not find biomarkers. &#8220People misunderstand that this is some type of blackbox technique for finding biomarkers.&#8220

Dr. Shockcor says the company is attempting to tune its UPLC systems to tackle the challenge of lipids since they are involved in many metabolic disorders, such as heart disease and diabetes. &#8220Were trying to make separation of lipids before they go through mass spectroscopy as simple as possible.&#8220

Thermo Electron ( is also advancing its MS technology. The LTQ Linear Ion Trap MS has three unique tools. One is its segmented ion trap where the center segment provides the purest field. &#8220A lot of analyzers produce a fringe field, which isnt good,&#8220 says Diane Cho, ion trap marketing manager. &#8220You want to create the most homogeneous field to trap and analyze your ions.&#8220 In addition, the ions are radially ejected into a dual detector. &#8220This allows us to detect all the ions with no loss.&#8220

The LTQ can be combined to high-resolution mass hybrids, the LTQ FT, and the LTQ OrbiTrap. &#8220These provide high-resolution and accurate mass assignment on a chromatographic-scale without having to use internal standards,&#8220 says Cho.

The system also provides cycle time to achieve high-throughput and high-sensitivity. &#8220We have the highest MSn spectral quality,&#8220 she states. &#8220We have customers using this for MS to the tenth. Its important because you need to look at the fragmentation pathway of unknowns to build back to your metabolite.&#8220 The system can also handle simultaneous quantitative and qualitative screening in one run.

Industry-Scale Metabolite Profiling

Metanomics Health ( uses 70 mass spectrometers along with its software for various metabolite applications. &#8220Our profiling platform focuses very wide into the metabolome&#8212we are picking up several thousands of metabolites in each sample,&#8220 says Arno Krotzky, Ph.D., CEO. The platform has been used in all steps of drug development, from toxicology to mode of action and into clinical stages.

&#8220What sets us apart from other approaches is that we are not looking at patterns. We are looking at analytes and allocating them to pathways.&#8220 This is accomplished by the companys tool for pathway visualization, Pathway Explorer. Dr. Krotzky explains that they incorporate literature knowledge and use data from their database, MetaMap, in which they enter metabolic profiling data, phenotypic data, clinical data, etc.

The final result data is projected into the Pathway Explorer. A graphical user interface helps scientists to navigate through the data and explore the impact of different drugs on specific parts of metabolism. An automated data validation tool integrates different types of data sets like gene expression and clinical data and applies a large suite of tools for statistical analysis.

Tracing Metabolic Reactions

SiDMAP (stable isotope-based dynamic metabolic profiling; uses its tracer technology to measure metabolic pathway flux to gain insight into cell function. &#8220Tracer-based metabolomics is a new field,&#8220 says Laszlo Boros, M.D., CSO. &#8220The advantage is that you can look dynamically at what happens to your substrate and what type of products the cell or the phenotype will generate from them.&#8220 This differs from traditional metabolic profiling since it doesnt measure overall metabolite levels or search for a single metabolite.

A stable isotope, like carbon-13, is incorporated into glucose. Mass spectrometry detects the different atomic numbers and shows the positions the tracer isotopes assume in downstream metabolites. &#8220We use the tracer and from the label patterns, we reconstruct the cells metabolic network and find specific reactions that best characterize the phenotype,&#8220 Dr. Boros explains. Many pathways in many different cells can be measured in one experiment, including glycolysis, fatty acid synthesis, pentose cycle, direct glucose oxidation, and the TCA cycle.

Applications include drug targets and effects, toxicity, and disease mechanisms. Potential future applications include optimizing patient populations for clinical trials and a primary drug development tool for cancer. &#8220Using our tracers, we can detect metabolites in very small amounts of plasma, which can be a screening tool for patients predisposed to an underlying metabolic abnormality or disease,&#8220 says Dr. Boros.

Predictive Models

Bio-Modeling Systems ( developed an approach to discover pathways or mechanisms involved in development of a disorder or pathology. &#8220These are qualitative models that take into consideration genetics, biochemistry, and physiology of the system as a whole,&#8220 says Francois Iris, Ph.D., company founder and CSO. &#8220We must be able to build a model that includes many cell types of a certain tissue. What sort of cross-talk goes on between those particular cell types in the healthy state and how does that differ from the development of pathology?&#8220 This, he says, provides a potential target for therapeutic effects.

Models are built using public information and proceed using a hypothesis. Literature is used to demonstrate if a hypothesis is correct. &#8220A model is not supposed to be a true representation of reality; its only an approximation,&#8220 says Dr. Iris. Models are tested biologically for validation. &#8220But now there is a big advantage, because I know what to look for, where, when, how, and most importantly, why.&#8220

New experimental results are put back into the modeling process and this gives rise to a new version of the model. Dr. Iris says it usually takes two to three rounds of experiments before one has a model where the major predictions are found to be biologically true.

The company has published a breast cancer model and presented and validated a Creutzfeld-Jakob disease model. A chronic fatigue syndrome model is currently under development.

Previous articleDr. Leroy Hood M.D. Co-founder of the Institute for Systems Biology in Seattle, Washington talks with GEN Editor-in-Chief John Sterling about Systems Biology.
Next articleDr. Raju Kucherlapati Scientific Director, Harvard-Partners Center for Genetics and Genomics & Paul Cabot Professor of Genetics, Harvard Medical School