Correlation-Based Deconvolution (CorrDec) To Generate High-Quality MS2 Spectra from Data-Independent Acquisition in Multisample Studies

Abstract

Data-independent acquisition mass spectrometry (DIA-MS) is essential for information-rich spectral annotations in untargeted metabolomics. However, the acquired MS2 spectra are highly complex, posing significant annotation challenges. We have developed a correlation-based deconvolution (CorrDec) method that uses ion abundance correlations in multisample studies using DIA-MS as an update of our MS-DIAL software. CorrDec is based on the assumption that peak intensities of precursor and fragment ions correlate across samples and exploits this quantitative information to deconvolute complex DIA spectra. CorrDec clearly improved deconvolution of the original MS-DIAL deconvolution method (MS2Dec) in a dilution series of chemical standards and a 224-sample urinary metabolomics study. The primary advantage of CorrDec over MS2Dec is the ability to discriminate coeluting low-abundance compounds. CorrDec requires the measurement of multiple samples to successfully deconvolute DIA spectra; however, our randomized assessment demonstrated that CorrDec can contribute to studies with as few as 10 unique samples. The presented methodology improves compound annotation and identification in multisample studies and will be useful for applications in large cohort studies.

Creating a Reliable Mass Spectral–Retention Time Library for All Ion Fragmentation-Based Metabolomics

“…To produce a clean MS2 spectrum using MS2Dec [9], an appropriate amount for each compound should be injected into the LC–MS system. CorrDec [11] requires multiple samples, with varying levels of the target compound [11]. Therefore, multiple injections at different dilutions are necessary.…”

“…We used two deconvolution methods based on different concepts. MS2Dec [9] applies a least square regression method to consider the difference of liquid chromatographic peak tops, while CorrDec [11] calculates the Pearson’s correlation among multiple samples to identify correlated MS2 peaks with the precursor. In other words, MS2Dec and CorrDec consider different information: ion intensity over RT in MS2Dec, and ion intensity across samples in CorrDec.…”

“…DIA-based data such as AIF (all ion fragmentation), MS E , or SWATH (sequential windowed acquisition of all theoretical fragment ion mass spectra) [8] are therefore rich in content, but require spectral deconvolution. Towards this end, multiple software programs such as MS2Dec [9], MetDIA [10], and CorrDec [11] have been developed for interpretation of DIA-based data. In this process, there is little consensus on the treatment of spectra originating from identical compounds such as in-source fragmentation and different adducts [12].…”

Spatiotemporal Determination of Metabolite Activities in Multiple Corneal Epithelium Barriers on a Chip

“…The MS2 similarity was scored as a simple dot product without any weighting (at least two MS2 peaks matched with the reference spectra). The MS2 similarities with reference spectra were matched to any of the CorrDec [62] or MS2Dec [60] deconvoluted MS2 spectra of the three collision energies (0, 10, and 30 eV). Peak areas exported from MS-DIAL were used for metabolite semiquantification.…”

Identification of Food-Derived Isoflavone Sulfates as Inhibition Markers for Intestinal Breast Cancer Resistance Proteins

“…MS-DIAL software (version 4.38) (Tsugawa et al, 2019) was used to process the AIF-acquired data to detect and align peaks using the following parameters: minimum peak height for MS1, 1000; MS1 and MS2 tolerance, 0.01 Da; retention time tolerance for peak alignment, 0.3 min; MS1 tolerance for peak alignment, 0.1 Da; more than 2/3 of samples were evaluated in each group. MS/MS spectra were obtained by retention time-based (MS2Dec) (Tsugawa et al, 2019) and correlation-based (CorrDec) (Tada et al, 2020) deconvolution methods using the following parameters: minimum peak height for MS2 (1000) and minimum correlation coefficient (0.8). A multivariate data matrix containing information on sample identity, retention time, m/z, and peak intensities was generated.…”