|
|
ACD/Labs ASMS 2007 Seminar
Hyatt Regency, National Parks/Concept
Indianapolis, IN, USA
Sunday, June 3, 2007
One South Capitol Avenue
Indianapolis, IN, USA 46204
ACD/Labs' Poster Schedule
| |
WEDNESDAY, JUNE 6, 2007
|
| Title: | Rapid Metabolite Identification using Advanced Algorithms for Mass Spectral Interpretation |
| Authors: | Mark Bayliss, Margaret Antler, Graham McGibbon, Vitaly Lashin (ACD/Labs) |
| Time: | 10:30 am to 2:30 pm |
| Location: | Exhibit Hall |
| Abstract #: | 2678 |
| Abstract: | Novel Aspect: Application of multiple redundancy approach to extraction of relevant LC/MS features from metabolite identification studies.
Introduction: Recognizing differences between related LC/MS data sets is the basic premise for the determination of potential metabolites in drug development. Finding small differences between two or more datasets requires a deep and rigorous analysis of each data set to extract and determine those m/z values that give rise to chromatographic peaks. The major challenge is that signal-to-noise ratio decreases as the limit of detection is approached so the number of false positive peaks that populate the output increases exponentially. In searching for relevant and important potential metabolites it is critical that as many false positive potential metabolite features can be eliminated from the output to reduce the human investment of time and energy in reviewing the results.
Methods: Using a unique algorithm for componentization of LC/MS data, we have developed software that uses extracted ion chromatogram peaks and as many mass spectral identifiers as possible to ensure that all features relevant to a chemical species are identified. We assume that for a mass spectral isotope cluster to be relevant, it should contain at least two or more contributing isotopes which share the same chromatographic elution profile and peak shape. Further, when considering the isotopes within a cluster, the ratio of contributions of the 12C and 13C should be within acceptable pattern distributions based on the number of contributing carbons. These identifiers are termed "peak tags". Multiple peak tags are used to extract relevant chromatographic components from data sets.
Preliminary results: Using the multiple information redundancy approach to data extraction, we determined with good accuracy the important and relevant components within a series of data sets. This was achieved by applying a combination of peak tags to filter the data so that only relevant information remained, based on mass spectral characteristics. For example: peaks that are tagged with—12C (AND) 12C: 13C within acceptable limits (AND) Mass Defect is +/- (0.2 Da). Using this approach, data sets containing more than 10 000 extracted ion chromatogram peaks were reduced to 80–150 relevant chromatographic components. The resultant lists of potential components were further reduced by comparing known metabolic mass differences from the parent mass. This outperformed traditional software approaches to extracting potential metabolite peaks which rely only on the presence of a chromatographic peak for a particular mass, rather than considering all spectral characteristics. Comparison of a dosed sample with a control may be regarded as another data filter to extract metabolite signals. Using this approach, only components that differ significantly from the control were tagged as potential metabolite features, and further data reduction was possible. Chromatographic features such as peak intensity were compared by the software. Spectral features, such as the isotope pattern for each ion cluster were also compared by the algorithm Applying this approach, we were able to rapidly and reliably reduce complex data sets to just those components that were relevant. As each peak is tagged with multiple pieces of MS specific information, then this assists a reviewer of the spectra with important information about the nature of each mass within the spectrum. The 12C/13C, comparison with a control, and isotope pattern information are particularly important for reducing the number of false positive identifications without discarding relevant information (i.e. false negatives), even for ultra-trace components. |
|
| TOP |
This page was last updated
12 September 2007
|
