Best-Matched Internal Standard Normalization in Liquid Chromatography–Mass Spectrometry Metabolomics Applied to Environmental Samples

Abstract

The goal of metabolomics is to measure the entire range of small organic molecules in biological samples. In liquid chromatography-mass spectrometry-based metabolomics, formidable analytical challenges remain in removing the nonbiological factors that affect chromatographic peak areas. These factors include sample matrix-induced ion suppression, chromatographic quality, and analytical drift. The combination of these factors is referred to as obscuring variation. Some metabolomics samples can exhibit intense obscuring variation due to matrix-induced ion suppression, rendering large amounts of data unreliable and difficult to interpret. Existing normalization techniques have limited applicability to these sample types. Here we present a data normalization method to minimize the effects of obscuring variation. We normalize peak areas using a batch-specific normalization process, which matches measured metabolites with isotope-labeled internal standards that behave similarly during the analysis. This method, called best-matched internal standard (B-MIS) normalization, can be applied to targeted or untargeted metabolomics data sets and yields relative concentrations. We evaluate and demonstrate the utility of B-MIS normalization using marine environmental samples and laboratory grown cultures of phytoplankton. In untargeted analyses, B-MIS normalization allowed for inclusion of mass features in downstream analyses that would have been considered unreliable without normalization due to obscuring variation. B-MIS normalization for targeted or untargeted metabolomics is freely available at https://github.com/IngallsLabUW/B-MIS-normalization .

Use of exogenous glycine betaine and its precursor choline as osmoprotectants in Antarctic sea‐ice diatoms1

“…GBT and choline were quantified in the aqueous fraction by standard addition into a representative sample matrix and analyzed by liquid chromatography–mass spectrometry using a Waters Acquity I‐Class UPLC fitted with a SeQuant ZIC‐pHILIC column and coupled to a Waters Xevo TQ‐S triple quadrupole with electrospray ionization; separation and detection as in Boysen et al. ().…”

“…After 6 d, cells were collected on PTFE filters and stored at À80°C until extraction. Osmolytes were extracted using a modified Bligh-Dyer extraction (Bligh andDyer 1959, Canelas et al 2009) resulting in an aqueous and organic fraction as previously described (Boysen et al 2018). GBT and choline were quantified in the aqueous fraction by standard addition into a representative sample matrix and analyzed by liquid chromatography-mass spectrometry using a Waters Acquity I-Class UPLC fitted with a SeQuant ZIC-pHILIC column and coupled to a Waters Xevo TQ-S triple quadrupole with electrospray ionization; separation and detection as in Boysen et al (2018).…”

Analytical considerations for (un)‐targeted metabolomic studies with special focus on forensic applications

“…In classical analytical approaches isotope-labeled IS are used to minimize these variations. 48,49 Normalization of each metabolite to a single IS assumes that all compounds experience the same variations as the IS. We assigned an approach based on similarity in specific chemical properties (chemical structure, RT) between the analyte and the IS.…”

“…Normalization of all analytes to the same IS, 46 to several IS based on chemical similarity and RT 17,47 or application of an IS-based normalization process. 48,49 Normalization of each metabolite to a single IS assumes that all compounds experience the same variations as the IS. This application is not really feasible for untargeted metabolomics data.…”

Tools and resources for metabolomics research community: A 2017–2018 update

“…For instance, a recent study that investigated patterns of missing data in an MS‐based metabolomics experiment of serum samples from the German Cooperative Health Research in the region of Augsburg (KORA) S4/F4 cohort ( n = 1750) using 31 imputation methods revealed that k‐nearest neighbors (KNNs) imputation on observations with variable preselection showed robust performance across all evaluation schemes and was computationally more tractable . Also, newer strategies for normalization, such as best‐matched internal standard normalization , continue to emerge. Thus, we summarize the tools that appeared in the 2018 horizon.…”