An Online Peak Extraction Algorithm for Ion Mobility Spectrometry Data

Abstract

Ion mobility (IM) spectrometry (IMS), coupled with multi-capillary columns (MCCs), has been gaining importance for biotechnological and medical applications because of its ability to detect and quantify volatile organic compounds (VOC) at low concentrations in the air or in exhaled breath at ambient pressure and temperature. Ongoing miniaturization of spectrometers creates the need for reliable data analysis on-the-fly in small embedded low-power devices. We present the first fully automated online peak extraction method for MCC/IMS measurements consisting of several thousand individual spectra. Each individual spectrum is processed as it arrives, removing the need to store the measurement before starting the analysis, as is currently the state of the art. Thus the analysis device can be an inexpensive low-power system such as the Raspberry Pi. The key idea is to extract one-dimensional peak models (with four parameters) from each spectrum and then merge these into peak chains and finally two-dimensional peak models. We describe the different algorithmic steps in detail and evaluate the online method against state-of-the-art peak extraction methods.

Current breathomics—a review on data pre-processing techniques and machine learning in metabolomics breath analysis

“…The last step of data pre-processing consists of peak picking. Several automated strategies are available such as merged peak cluster localization [49], growing interval merging [46], and wavelet-based multiscale peak detection [46], watershed transformation (WST) [47] and peak model estimation (PME) [50]. We focus on the most widely used approaches for MCC-IMS data and give their key ideas in the following.…”

“…This approach uses continuous functions providing non-integer peak positions, which is presumably more precise than the discrete results of the previous methods. Kopczinski et al published an implementation of this approach [50]. It includes a seed-finding method based on finding roots in the first derivatives of both spectra and chromatograms [52].…”