Automated classification of diglucosides glycosidic linkage with ion mobility spectrometry data by machine learning approaches

Radecki, Michal (2020) Automated classification of diglucosides glycosidic linkage with ion mobility spectrometry data by machine learning approaches. Masters thesis, UNSPECIFIED.

[img]
Text (2020radeckimscbyresearch)
2020radeckimscbyresearch.pdf - Published Version
Available under License Creative Commons Attribution-NonCommercial-NoDerivs.

Download (1MB)

Abstract

Unambiguous characterization of carbohydrate products remains a challenging endeavour. The current state-of-the-art techniques are NMR-based approaches, which require large amounts of purified sample and are challenging to annotate. Recently, a strategy has been developed combining gas-phase ion-mobility spectrometry with tandem mass spectrometry to separate and characterize isomeric product ions. Crucially, this can provide information about the stereochemistry that MS alone is often “blind” to because certain monosaccharides have the same m/z (i.e. they are isomeric) and are therefore indistinguishable. Given the amount of data this approach produces, there is a need for a method to rapidly annotate the produced data. The initial strategy involves developing a method that can discern the number of peaks within an IMS spectrum, where it has been shown that product ions derived from α-glucosides produced similar features (2 peaks) whereas β-glucosides only produced a single peak. It was reported that an IMS signal can be approximated as a sum of spectral line shapes (such as Gaussian, Lorentzian or Voigt). Current results show that the approximation method allows analysis of the signal in terms of peaks description. Using this approach, we built a feature matrix from IMS data for different diglucosides, which was then used to train a machine learning classifier able to distinguish between α- and β-glucosides. The performance of the classifier proved that automated classification of glucosides by their bonding type is achievable.

Item Type:
Thesis (Masters)
ID Code:
173347
Deposited By:
Deposited On:
19 Jul 2022 08:30
Refereed?:
No
Published?:
Published
Last Modified:
21 Nov 2022 12:20