Prioritization of putative metabolite identifications in LC-MS/MS experiments using a computational pipeline

Abstract

One of the major bottle-necks in current LC-MS-based metabolomic investigations is metabolite identification. An often-used approach is to first look up metabolites from databases through peak mass, followed by verification of the obtained putative identifications using MS/MS data. However, the mass-based search may provide inappropriate putative identifications when the observed peak is from isotopes, fragments, or adducts. In addition, a large fraction of peaks is often left with multiple putative identifications. To differentiate these putative identifications, manual verification of metabolites through comparison between biological samples and authentic compounds is necessary. However, such experiments are laborious, especially when multiple putative identifications are encountered. It is desirable to use computational approaches to obtain more reliable putative identifications and prioritize them before performing experimental verification of the metabolites. In this article, a computational pipeline is proposed to assist metabolite identification with improved metabolome coverage and prioritization capability. Multiple publicly available software tools and databases, along with in-house developed algorithms, are utilized to fully exploit the information acquired from LC-MS/MS experiments. The pipeline is successfully applied to identify metabolites on the basis of LC-MS as well as MS/MS data. Using accurate masses, retention time values, MS/MS spectra, and metabolic pathways/networks, more appropriate putative identifications are retrieved and prioritized to guide subsequent metabolite verification experiments.

Figure 1

Proposed computational pipeline for metabolite identification and prioritization of putative identifications.

Figure 2

Reconstructed biochemical network based on putative identifications from the positive mode data (top panel) and the negative mode data (bottom panel). Each node represents a putative identification. Isolated putative identifications (i.e., those with no connection to others) are omitted from the figure. The enlarged nodes (marked in yellow) in the bottom panel are provided in Figure 3.

Figure 3

A network cluster derived from the pathway and network analysis result presented in Figure 2 (bottom panel). Each node is a peak group in LC-MS data, labeled by its putative identification with the highest probability from the pathway analysis. The nodes labeled as 3-Sulfo-GCDCA and TCA represent the most likely candidates for M4 and M5, respectively.

Figure 4

MS/MS spectral comparison of experimental samples M3, M4, and M5, versus authentic compounds S-1-P, 3-sulfo-GCDCA, and TCA.