qAlgorithms: A parameter-free, quality-based data processing routine for non-target-screening with high-resolution mass spectrometry.
Year: 2024
Type: dissertation
Abstract: This thesis focuses on the current challenges and limitations of data processing in non-target screening (NTS) using chromatographic methods coupled to high-resolution mass spectrometry. Due to the complexity and volume of NTS data, the use of automated processing routines is necessary, but established methods often have low transparency and do not provide information on the reliability of results due to variable user parameter inputs and the unclear effect of varying data quality. This thesis addresses these challenges by developing three innovative algorithms for NTS data processing that are meaningfully combined and integrated into a common toolbox (qAlgorithms). A central role of the algorithms is to investigate the uncertainty with which feature properties (including m/z, retention time, and peak area) are estimated in the mathematical processes of the algorithms. This information is then summarized for the user in the form of a Data Quality Score (DQS) and appended to the generated feature list. The first algorithm developed in this thesis (qCentroids) is used to centroid high-resolution profile mass spectra. It uses a log-transformed variant of the Gaussian peak function, which can be solved using linear algebra. In contrast to non-linear optimization methods for peak modeling, no initial parameters need to be set and the robustness of the results can be estimated from the standard errors of the regression coefficients. The generation of extracted ion chromatograms (EICs) from the previously calculated mass centroids was the focus of the second algorithm, called qBinning. In qBinning, m/z values are grouped using a statistical test derived from order statistics so that m/z originating from identical ion populations are assigned to each other. To evaluate the quality of the bins, a performance metric from cluster analysis, the silhouette criterion, has been adapted for this specific processing step. The last data processing step of the NTS covered in this thesis is the detection and characterization of chromatographic peaks within EICs (qPeaks). For this purpose, a new peak model for the description of asymmetric chromatographic peaks was developed which, in contrast to conventional methods, can be calculated using linear algebra. This means that it is independent of initial parameters and can be used in an algorithm for the combined detection and characterization of chromatographic peaks. The regression is scanned over the chromatogram and the regressions that describe meaningful and statistically significant peaks are stored. A subsequent grouping of non-differentiable peaks completes the algorithm. The qAlgorithms toolbox and all algorithms implemented in this work are available as open source to the scientific community to enable the broad and sustainable application of the concepts developed here. In particular, the concept of the DQS for evaluating the reliability of the collected processing results offers great potential for prioritizing and weighting NTS features in the comparative analysis of samples of different origins in the future.
Author: Reuschenbach, Max
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W6962939217
Cites:
Cited by:
Related to: