Local Asymmetric Gaussian Fitting Algorithm for Enhanced Peak Detection of Liquid Chromatography–High Resolution Mass Spectrometry Data

• Published in: Analytical chemistry (Washington) Vol. 97; no. 20; pp. 10603 - 10610
• Main Authors: Zou, Shengsi, Cui, Qingxiao, Liu, Jinyue, Wu, Qiong, Zhu, Lijia, Chen, Da, Du, Yiping, Wu, Ting
• Format: Journal Article
• Language: English
• Published: United States American Chemical Society 27.05.2025
• Subjects: Algorithms; analytical chemistry; Asymmetry; blood serum; Chromatography; computer software; Computing time; Data points; Goodness of fit; hydrophilicity; Liquid chromatography; Mass spectrometry; Mass spectroscopy; Metabolomics; Parameters; Python; Scientific imaging; Source code; Standard deviation
• ISSN: 0003-2700; 1520-6882; 1520-6882
• DOI: 10.1021/acs.analchem.5c00060

Feature detection is a crucial step in the data preprocessing workflow of liquid chromatography–mass spectrometry (LC–MS). However, many existing methods are hindered by intricate parameter adjustments and high false positive rates during extracted ion chromatogram (EIC) construction and peak detection, which challenges the identification of spurious and missing compounds. This study introduces a novel algorithm, local asymmetric Gaussian fitting (LAGF), for peak detection. LAGF integrates with the “data points bins” EIC extraction algorithm to enhance the feature detection efficiency. By using a 1 Da data points bin for EIC extraction, computational time is significantly reduced, making the method well-suited for batch metabolomics analysis. LAGF minimizes parameter numbers of generalized two-sided asymmetric Gaussian fitting by automatically determining the peak center (μ) and height (α) while accommodating two-sided standard deviations (σ1 and σ2) to self-adaptively model peak patterns. Features are filtered based on a goodness-of-fit threshold of 0.5. The performance of LAGF was validated using standard mixtures and serum samples at different concentrations in reversed-phase or hydrophilic interaction LC mode. In most cases, LAGF outperformed conventional tools in terms of determination coefficient (R 2) and relative standard deviation for automatically detected peak areas. The LAGF algorithm is available as open-source Python code alongside an interactive interface, facilitating implementation in both nontargeted and targeted LC–MS analysis to enhance peak detection and compound identification.