Prioritisation algorithms for data acquisition in liquid chromatography mass spectrometry

McBride, Ross Daniel (2025) Prioritisation algorithms for data acquisition in liquid chromatography mass spectrometry. PhD thesis, University of Glasgow.

Abstract

Liquid chomatography mass spectrometry (LC-MS/MS) is a powerful analytical platform frequently used to identify the composition of biological samples. For example, LC-MS/MS is one of the leading measurement technologies within metabolomics, which has applications in discovering disease biomarkers and novel drugs, in ecology and environmental science and in forensics and toxicology, among many others.

The goal of an untargeted LC-MS/MS experiment is to discover as many unique analytes in the sample as possible in order to generate hypotheses relevant to the experiment’s goals. One of the most powerful tools in annotating analytes is the fragmentation spectra produced by tandem mass spectrometry, which are a kind of “molecular fingerprint” which can be matched against databases. However, collection of unambiguous fragmentation spectra requires individually targeting analytes for acquisition. As a consequence, resources (tandem mass spectrometry scans) must be efficiently allocated in order to collect as many fragmentation spectra as possible at the highest possible quality. The goal is to target as many possible “peaks” at the correct acquisition time to maximise their “intensity” (a proxy for acquisition quality).

To address this important resource allocation problem, this thesis presents several new “fragmentation strategies”. Firstly we present TopNEXt, a framework for Data-Dependent Acquisition (DDA) strategies which utilises area and intensity comparisons between LC-MS/MS runs to develop advanced DDA strategies. We show that the strategy using all of these features, Intensity Non-Overlap is highly effective and is able to acquire fragmentation spectra for an additional 10% of our set of target peaks and with an additional 20% of acquisition intensity.

We then present a “pre-scheduled” method which uses a maximum bipartite matching algorithm to plan an acquisition in advance. We extend an existing technique to map the LC-MS/MS acquisition problem to an instance of the maximum bipartite matching problem. Our extensions include extending the technique to plan multiple runs and samples as a set, solving a weighted version of the problem to optimise acquisition times and redundantly assigning unassigned scans to improve the robustness of the method. We show that this schedule can theoretically obtain completely comprehensive coverage of a sample in a low number of injections compared to other methods. However, we also investigate the trade-off between DDA and pre-scheduled methods by testing this pre-scheduled method in a situation significantly different than the one which it has planned for (which may happen frequently in reality). In this scenario we show that it still has performance comparable to the state-of-the-art, but only with the improvements we have made to the technique. Finally, we reflect on the common elements that make our techniques successful: namely, accounting for acquisition time and quality, and judicious use of redundancy to improve their robustness.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
• Colleges/Schools: College of Science and Engineering > School of Computing Science
• Supervisor's Name: Bryson, Dr. Kevin
• Date of Award: 2025
• Depositing User: Theses Team
• Unique ID: glathesis:2025-85035
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 10 Apr 2025 13:47
• Last Modified: 10 Apr 2025 14:16
• Thesis DOI: 10.5525/gla.thesis.85035
• URI: https://theses.gla.ac.uk/id/eprint/85035