Novel probabilistic similarity scores for sets of replicate EI mass spectra

Published

March 5, 2025

Author(s)

Amudhan Krishnaswamy Usha, Briana Capistran, Anthony Kearsley

Abstract

Abstract: Background: Mass spectra are common signatures used to discriminate between compounds. This often involves the use of a similarity score to classify and distinguish between spectra of different compounds. To separate spectra of structurally similar compounds, multiple authors have explored the use of statistical and probabilistic methods applied to replicate mass spectra. In this paper, we explore the use of various averaged versions of the Kolmogorov–Smirnov and t-test statistics to compare peak intensities for sets of replicate mass spectra. Results: Using replicate gas chromatography electron ionization (GC-EI-MS) mass spectra of 25 forensically relevant compounds, we compare the 'library match' and total classification accuracy of our novel probabilistic similarity scores versus the 'high dimensional consensus' (HDC) score and other similarity scores previously used in this context. We show that the use of the harmonic mean of the Kolmogorov–Smirnov test statistics obtained from peak intensities results in accuracies comparable to the HDC score. Significance: Our results provide novel probabilistic similarity scores for replicate EI mass spectra, which outperform traditional scoring methods while at the same time minimizing the number of user-defined parameters and avoiding assumptions about the distributions of peak intensities.

Citation

Analytica Chimica ACTA

Volume

1351

Pub Type

Journals

Download Paper

https://doi.org/10.1016/j.aca.2025.343858.

Local Download

Keywords

Mass spectral similarity, Mass spectrometry, High dimensional consensus spectra, Kolmogorov–Smirnov scores

Citation

Krishnaswamy Usha, A. , Capistran, B. and Kearsley, A. (2025), Novel probabilistic similarity scores for sets of replicate EI mass spectra, Analytica Chimica ACTA, [online], https://doi.org/10.1016/j.aca.2025.343858., https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=958661 (Accessed March 19, 2025)