Washington University in St. Louis

The Patti Lab
Metabolomics to elucidate novel biochemical mechanisms of disease
Job Opportunities
Job opportunities are available.

Discriminating Precursors of Common Fragments for Large-Scale Metabolite Profiling by Triple Quadrupole Mass Spectrometry

Nikolskiy I, Siuzdak G, Patti GJ
Discriminating Precursors of Common Fragments for Large-Scale Metabolite Profiling by Triple Quadrupole Mass Spectrometry
Bioinformatics, 31(12), 2017-2023, 2015
doi:10.1093/bioinformatics/btv085

Motivation: The goal of large-scale metabolite profiling is to compare the relative concentrations of as many metabolites extracted from biological samples as possible. This is typically accomplished by measuring the abundances of thousands of ions with high-resolution and high mass accuracy mass spectrometers. Although the data from these instruments provide a comprehensive fingerprint of each sample, identifying the structures of the thousands of detected ions is still challenging and time intensive. An alternative, less-comprehensive approach is to use triple quadrupole (QqQ) mass spectrometry to analyze predetermined sets of metabolites (typically fewer than several hundred). This is done using authentic standards to develop QqQ experiments that specifically detect only the targeted metabolites, with the advantage that the need for ion identification after profiling is eliminated.

Results: Here, we propose a framework to extend the application of QqQ mass spectrometers to large-scale metabolite profiling. We aim to provide a foundation for designing QqQ multiple reaction monitoring (MRM) experiments for each of the 82‚ÄČ696 metabolites in the METLIN metabolite database. First, we identify common fragmentation products from the experimental fragmentation data in METLIN. Then, we model the likelihoods of each precursor structure in METLIN producing each common fragmentation product. With these likelihood estimates, we select ensembles of common fragmentation products that minimize our uncertainty about metabolite identities. We demonstrate encouraging performance and, based on our results, we suggest how our method can be integrated with future work to develop large-scale MRM experiments.

Availability and implementation: Our predictions, Supplementary results, and the code for estimating likelihoods and selecting ensembles of fragmentation reactions are made available on the lab website at http://pattilab.wustl.edu/FragPred.

Washington University, Departments of Chemistry, Genetics, and Medicine. Saint Louis, Missouri 63110 USA