Mass Difference Maps and Their Application for the Recalibration of Mass Spectrometric Data in Nontargeted Metabolomics

Abstract

Modern high-resolution mass spectrometry provides the great potential to analyze exact masses of thousands of molecules in one run. In addition, the high instrumental mass accuracy allows for highprecision formula assignments narrowing down tremendously the chemical space of unknown compounds. The adequate values for a mass accuracy are normally achieved by a proper calibration procedure that usually implies using known internal or external standards. This approach might not always be sufficient in cases when systematic error is highly prevalent. Therefore, additional recalibration steps are required. In this work, the concept of mass difference maps (MDiMs) is introduced with a focus on the visualization and investigation of all the pairwise differences between considered masses. Given an adequate reference list of sufficient size, MDiMs can facilitate the detection of a systematic error component. Such a property can be potentially applied for spectral recalibration. Consequently, a novel approach to describe the process of the correction of experimentally derived masses is presented. The method is based on the estimation of the density of data points on MDiMs using Gaussian kernels followed by a curve fitting with an adapted version of the particle swarm optimization algorithm. The described recalibration procedure is examined on simulated as well as real mass spectrometric data. For the latter case, blood plasma samples were analyzed by Fourier transform ion cyclotron resonance mass spectrometry. Nevertheless, due to its inherent flexibility, the method can be easily extended to other low-and high-resolution platforms and/or sample types.

Decomposing the molecular complexity of brewing

“…The used reagents, sample preparation, and instrumental parameters are given in Supplementary Table 7. Possible space charge effects were recalibrated by mass difference mapping 44 . The samples were measured over a period of 18 months in randomized order using a representative lager beer as quality control.…”

Linking the FTO obesity rs1421085 variant circuitry to cellular, metabolic, and organismal phenotypes in vivo

“…Gibbs peaks were removed by Data Analysis automatically. All spectra were exported as tab-separated asc files and loaded into the Kernel Calibrator ( 27 ). A mouse calibrant list was generated manually after first precalibrating representative mouse tissue spectra against fatty acids, followed by isotopic pattern simulation of the most abundant MS features.…”

“…Gibbs peaks were removed by Data Analysis automatically. All spectra were exported as tab-separated asc files and loaded into the Kernel Calibrator ( 27 ). m/z calibration was then performed against an in-house repository of human blood plasma metabolites for both cohorts.…”

“…4A). We used mass difference network (MDiN) analysis to assign MS features to high-confidence molecular formulas encompassing known and unknown metabolites (23)(24)(25)(26)(27). Here, "known metabolic features" refers to molecular formulas that have a match in the Human Metabolome database (HMDB) database (28).…”

Mass recalibration for desorption electrospray ionization mass spectrometry imaging using endogenous reference ions

“…Subsequently, we further filter the list of candidate reference ions using a method based on a kernel density estimator (KDE) following a similar procedure to that described in Smirnov et al (Smirnov et al, 2019) If candidate reference peaks for M are represented as points with coordinates , we assume that highly dense connected regions represent the shift trends of M .…”