|
«
|
November
2009
|
»
|
| Su |
Mo |
Tu |
We |
Th |
Fr |
Sa |
| 1 |
2 |
3 |
4 |
5 |
6 |
7 |
| 8 |
9 |
10 |
11 |
12 |
13 |
14 |
| 15 |
16 |
17 |
18 |
19 |
20 |
21 |
| 22 |
23 |
24 |
25 |
26 |
27 |
28 |
| 29 |
30 |
|
|
|
|
|
|
ASMS 2009
| Code: Wednesday Oral - WOB am |
|
|
| Time Slot/Poster Number: 08:30 am |
Session: Advances in Metabolomics |
Metabolic signatures of human drug response phenotyping
|
| Oliver Fiehn1; Gert Wohlgemuth1; Dinesh Kumar Barupal1; Tobias Kind1; Rima Kaddurah-Daouk2 |
1UC Davis, Davis , CA; 2Duke University, Durham, NC
|
| Novel Aspect: |
Personalized medicine by a time-of-flight MS database approach
|
| Introduction: |
Prescribed pharmaceutical drugs are only effective in sub-cohorts of the general population, but not in all patients who are medicated. Current diagnostics fail to accurately predict the individuals who are going to benefit most from a specific agent, who would likely not see effects and who would be better treated by an different medicine. SNP genotyping is a promising tool towards personalized medicine, but it can only be used for general genetic predispositions for diseases nd drug efficacy. In order to pinpoint the current state of an individual, including metabolic absorption and turnover rates, mass spectrometry-based metabolomics can be used to discern metabolism before and after drug treatment.
|
| Methods: |
EDTA-blood plasma was used from patient cohorts in a variety of drug response phenotype studies for comprehensive metabolite extraction by cold protein precipitation at -20C and iPrOH/ACN/H2O (vol.ratio 30:1, mixture 5/2/2). After trimethylsilylation, GC-TOF mass spectrometry (Leco Pegasus IV) was performed using Gerstel automatic liner exchange and cold injection conditions. GC-TOF spectra were filtered by the in-house BinBase mass spectral database. Metabolites were identified by a retention index/MS library of 713 authentic standards including PubChem and KEGG identifiers. Quantitative results were statistically evaluated, and significant differences were mapped to biochemical and chemical databases by open-access tools. Unknown metabolites were annotated using chemical ionization GC-TOF MS (Waters GC-T) and database queries according to accurate mass and accurate isotope data.
|
| Preliminary Results: |
Patient cohorts were divided into the upper and lower 10% of individuals who responded well or poorly to a pharmaceutical treatment. After sample preparation, EDTA-plasma GC-TOF chromatograms showed on average over 650 deconvoluted peaks. In order to reduce the data set to peaks that are consistently detectable in subsequent analysis of biological replicate samples, the in-house database BinBase was employed by using mass spectral metadata such as peak purity, unique masses, s/n and apex masses. Around half of the peaks did not match the various quality thresholds implemented in the database schema. The remaining 300+ compounds were queried for compound identification using the Agilent and Leco GC/MS libraries that combine both retention index and mass spectral match criteria for unambiguous metabolite identification. Pharmacogenomic studies yield over 120 identified compounds in this manner, automatically excluding known artifacts such as phthalates or column bleed polysiloxanes. Unidentified spectra were queried against the full BinBase database of more than 4,500 unique metabolite peaks using a web interface and yielded the top-10 hits for each compound in order to obtain substructure similarities for these spectra. Select unknowns were further investigated by accurate mass GCT analysis. Subsequently, quantitative results were analyzed for each study by multivariate and univariate statistics. In several cases it was shown that definite metabolic pathways contributed to the drug response phenotype for the studied patient cohort, suggesting a role of the gut microbiome in the overall response. In addition, data were visualized by chemical similarity networks (Tanimoto distances) to enable straightforward interpretations of major metabolic disruptions. Furthermore, efforts are represented to include distance metrics for spectra of unknown peaks in such networks by combining retention index and spectral match matrices into graph representations. This research was funded by NIH grants 5R01ES13932 and 4R24GM078233.
|
|
| Code: Wednesday Poster - WPD |
|
|
| Time Slot/Poster Number: 123 |
Session: Metabolomics II |
Metabolite profiling of the novel NIST standard human plasma using a multi-target calibration approach in GC/quadrupole MS
|
| Mine Palazoglu; Sevini Shahbaz; Oliver Fiehn |
UC Davis, Davis , CA
|
| Novel Aspect: |
Assessing routine metabolomics for the novel plasma standard (NIST/NIDDK) using quadrupole GC/MS.
|
| Introduction: |
Quadrupole mass spectrometers are used in a wide range of laboratory settings. With recent advances in data analysis techniques (AMDIS) and retention time locking software (Agilent), these affordable instruments offer new opportunities in the field of metabolomics. New developments enable continuous switching from single monitoring to scan mode in order to selectively detect a list of pre-defined analytes in high sensitivity while also providing data for spectral matching against established mass spectral libraries. This study shows how today’s upgraded GC/quadrupole mass spectrometers can be used in routine profiling of metabolites in biological matrices.
|
| Methods: |
Novel standard human plasma samples were obtained from NIST for certification purposes. 15 ul plasma aliquots were prepared for comprehensive metabolite extraction by cold protein precipitation at -20C and iPrOH/ACN/H2O (vol.ratio 30:1, mixture v/v/v 5/2/2). Stock solutions of calibration standards were prepared as 1 mg/ml in methanol-water and isopropanol mixture. Samples and standards were derivatized by methoximation and trimethylsilylation. Seven-point calibration curves were established at 50µg/ml-0.5µg/ml concentration ranges. An Agilent 6890 GC/5975MSD mass spectrometer was equipped with a DB5-MS +10m duraguard 30 m x 250 µm x 0.25 µm capillary column. Injector temperature was set at 250C with an oven program ramp from 60-325C. Masses were acquired from m/z 50 to 600.
|
| Preliminary Results: |
Chromatograms of replicate standard human plasma were first processed for identification of detectable metabolites, and 75 compounds were unambiguously assigned by AMDIS deconvolution using the Fiehn/Agilent metabolomics library. Subsequently, calibration curves of metabolite standards were established for 100 common metabolites and repeated three times for assessing stability of mass spectrometer sensitivity over periods of use. Spectra and locked retention times were located in a quantification target list for selective monitoring in a routine manner. An aggregated k-means clustering method was used to assign each metabolite to a different cluster of compounds based on Euclidean distances of retention times, logP, molecular weights and mass spectral sensitivity. This approach led to the selection of 29 of these compounds as quality control mixture for daily evaluation of instrument sensitivity to maintain in-control operation. The clustering approach ensured that these QC compounds were structurally diverse, biologically relevant and covered compounds of both high and low mass spectral response factors. This procedure also ensured that these QC calibration curves can be extended to semi-quantify metabolites that are added later to the profiling lists. Results are presented on the quantification of metabolites in standard human plasma under routine conditions.
|
|
| Code: Wednesday Poster - WPD |
|
|
| Time Slot/Poster Number: 125 |
Session: Metabolomics II |
Metabolomics of volatile compounds by a new BinBase mass spectral database
|
| Gert Wohlgemuth; Kirsten Skogerson; Oliver Fiehn |
UC Davis, Davis , CA
|
| Novel Aspect: |
Development of a novel database for tracking and identifying volatile compounds in complex and diverse mixtures.
|
| Introduction: |
Plants interact with the environment through the emission of volatile compounds. Over 1,700 volatiles from 90 different plant families have been identified. Biotic and abiotic factors affect the identity, quantity and timing of their release. Improving methods for capturing, cataloging and tracking volatile compounds from a wide range of plant families is required to be able to compare data across studies.
GC-TOFMS yields hundreds of metabolic signals per chromatogram. Existing solutions for peak-picking and data alignment rely on similarity matching across all samples of an experimental batch. This approach fails if highly dissimilar chromatograms are to be compared. The algorithm in the BinBase DB resolves these issues by stringent quality criteria.
|
| Methods: |
All data used in database development was collected on a Leco Pegasus GC-TOF-MS equipped with a Gerstel TDU/CIS4 injector. To facilitate automated annotation and database construction, all samples were spiked with a mixture of internal retention index (RI) markers (fatty acide methyl esters C4-C24) prior to analysis. Database construction implemented the same algorithms as established in the Fiehn laboratory for liquid injections of TMS-derivatized metabolite profiles. Adaptation of the existing framework for volatiles required optimization of existing algorithms and extension of the mass spectra range from 85-500 m/z to 0-1000 m/z. Additionally, the volatile database was created as a plug-in for the existing BinBase, so that both liquid and volatile sample types are calculated on a single version.
|
| Preliminary Results: |
Absolute retention times (RT) of GC/MS peaks shift by column age and column length, and small RT differences are also observed from sample to sample. Volatile retention index marker compounds (RI) are here introduced for the first time to lock the retention times of eluted compounds to fixed retention index positions. The BinBase algorithm starts with finding and validating RI compound spectra for a polynomial regression curve. Only if all RI markers are found for low- and high boiling compounds, new spectra (‘Bins’) are allowed to enter the DB.
Robust protocols for the introduction of RI markers into the volatile samples were developed using a FAME mixture from C4-C24. A variety of new algorithms had to be programmed to adjust the existing metabolomics BinBase database (version 3.3) from TMS-derivatized metabolites to the volatile Binbase database plugin. Parameters were changed to RI markers and to find and create new Bins. Matching settings and base peak patterns were changed as now spectra are queried from 0-1000 m/z. Parameters were validated by recalculating known test datasets of volatile emissions from citrus plants. Over the course of setup and testing of the volatile database 18,000 test calculations were performed on a series of 1,200 volatile samples collected in grapevine canopies in a study on grape maturation. A total of 796 Bins were generated which are partly annotated by matching both mass spectra and retention indices to a commercial library of 2,000 volatile spectra. All GC/TOF spectra, raw data, processed data and the BinBase source code are available for downloads.
|
|
| Code: Thursday Poster - ThPC |
|
|
| Time Slot/Poster Number: 073 |
Session: Metabolomics III |
Visualization of identified and unknown compounds in metabolomic data sets of environmental tobacco smoke exposure in rats
|
| Dinesh Kumar Barupal; Oliver Fiehn |
UC Davis, Davis , CA
|
| Novel Aspect: |
Applications of enhanced visualization techniques for metabolomics research in communication with collaborators in biology.
|
| Introduction: |
Mass spectrometry-based metabolomics studies generate multidimensional quantitative data of the differential regulation of metabolites in response to genetic, environmental or physiological perturbations. Biomarker discovery, understanding metabolic alterations, phenotypic discrimination and displaying relationships within such data sets require optimized visualizations. Result graphs need to utilize the mass spectrometric and statistical properties of the detected metabolic signals to highlight the experimental outcome for collaborators in biology.
|
| Methods: |
The BinBase database approach was used to filter inconsistent peaks from 700 signals per chromatogram to over 300 consistent metabolites per GC/TOF MS based studies on metabolic changes after exposure of rats to environmental tobacco smoke. Metabolites were identified by the Agilent and Leco MS libraries based on MS matches and retention indices. Unidentified metabolites were were subjected to multivariate statistics, clustering and visualization using open-access, and commercial software based on their MS similarity and RI differences to the grid of known compounds. Different visual properties of plots, networks and heatmaps were adjusted according to the ttest results of second hand smoke stress studies in rats.
|
| Preliminary Results: |
Using information visualization tools, we undertook three approaches to visualize the metabolic signals and ttest statistics: (A) networks (B) multivariate statistics visualization and (C) specialized graphs. Our common aim was to represent every detected metabolite; specifically, to highlight the significantly altered metabolites on graphs comprising a structured organization that represents biochemical or chemical similarities such as substructures.
Networks displaying the spectral similarity (NIST MS search algorithm) and chemical similarity (Pubchem-Tanimoto chemical similarity) were superior to all other visualizations in terms of organization, visual appearance, clarity and highlighting ttest statistics. However, Tanimoto networks can only be used for identified compounds, not for unknown peaks.
Therefore, NIST MSP format mass spectra of all detected metabolic signals were converted into a matrix for subsequent dimension reduction in PCA and PLS analysis. We projected the first two principal components in three-dimension space and plotted in 3D bubble plots. Resulting plots were found to be poorly structured due to the large degree of variation in metabolite spectra. Therefore, we combined the data mining techniques such as k-mean clustering with multivariate statistics for a better visualization. This joint approach enhanced the organization of the metabolites in 3D space. The resulting order of metabolic signals was plotted against retention indices in bubble charts. In a further study, multivariate statistics and data mining results were plotted against retention indices in motion charts that display the ttest results as an animation which was found to be useful to assist understanding time series experiments. In addition to dimension reduction and visualization, we used specialized graphs such as polar bubble charts, color-scale column charts and heatmap/treemap visualization techniques for highlighting the statistical results.
Results of all these approaches are contrasted and discussed for use in interactions with biologists.
|
|
Created by
zwluxx
Last modified
2009-04-16 08:11 PM
|
ASMS 2009
