Two-dimensional alignment of differential mobility spectrometer data

https://doi.org/10.1016/j.snb.2005.12.058Get rights and content

Abstract

Alignment of data is critical for analysis involving the comparison of multiple files. Analytical sensor data resulting from complex chemical mixtures can often be mis-aligned due to time-varying biases resulting from mechanical instrument variability. In addition to the necessity for time alignment, data from the micromachined differential mobility spectrometer (DMS) may be slightly shifted when comparing various data sets due to the effect of heat and flow variations on the compensation voltage (Vc). Thus, the data in this dimension can also benefit from alignment. We present here a method for the alignment of both dimensions (scans and Vc) of pyrolysis-DMS data, using a single file as reference. The Vc dimension is first aligned with respect to the reference file; this is a rigid shift and no interpolation is performed. This is an advantage as the Vc dimension has physical meaning and should not be altered by interpolation. The time (or scans) dimension is then aligned with respect to the reference by identifying common landmarks and interpolating according to a piecewise linear function calculated based on the amount of shift between the two files. The effect of a slight change in flow in the Vc dimension is examined using the nitrogen reactant ion peaks as a standard signal. This method is useful for further data processing in which multiple files are to be directly compared, and it could also be useful for two-dimensional alignment of data from other sensor modalities.

Introduction

Sensor output data structures can often be misaligned due to experimental variations such as heat and gas flow, and this misalignment can also make it more difficult to directly compare data collected over time. Misalignment of analytical sensor data in the time dimension, especially for chemical mixtures that are separated by gas chromatography (GC), is well-documented [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14]. In addition to the need for time alignment, we have also found that differential mobility spectrometry data resulting from complex chemical mixtures can often require alignment in the compensation voltage (Vc) dimension as well. The operation of the differential mobility spectrometer (DMS) has been described previously for various applications [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25]. The compensation voltage can be affected by the gas flow rate and moisture content of the sample as it passes through the DMS, as demonstrated previously by Miller et al. [26] and Krylova et al. [27] Small fluctuations in the flow rate or moisture content thus affect the Vc dimension in the data, and these can be easily standardized across all data sets by alignment.

Although much work has been done with alignment of data in a single dimension, there are relatively few methods published for aligning analytical sensor data from the detection of complex chemical mixtures in two dimensions. Advancements have been made in the field of image registration, where images are stretched, compressed, or rotated to match a reference image; several recent publications provide a comprehensive review of the methods used in this field [28], [29], [30], [31]. However, the considerations for alignment of data from an analytical sensor are often different than for image alignment, as the way in which the data is shifted can dramatically affect the analysis of components present. The alignment method must be sensitive to the physical properties of this data and cannot dilate, compress, or rotate the data simply to achieve the best overlap because the properties of the data may then be distorted. For data resulting from DMS analysis, the Vc at which a compound elutes is unique to that compound, based on its chemical and physical properties. Therefore, that dimension must not be distorted (dilated or compressed) because the physical meaning of the data would be lost. The alignment of this data thus requires a rigid shift in the Vc dimension due to linear changes in instrument detection. In the time dimension more flexibility can be allowed for the axis to be compressed or dilated as necessary due to changes in elution time that may result from the separation step prior to detection. For this reason, we decided not to pursue image alignment methods, but rather to create a new two-dimensional (2D) alignment method that would provide us with these requirements.

In the analytical chemistry field, one group has published several articles regarding alignment of comprehensive two-dimensional gas chromatography (GC × GC) [32], [33], [34], [35], [36]. In this method, a sample run is compared to a standard, and usually only a small part of a chromatogram is compared at a time. The peaks are considered to be perfectly aligned when the pseudorank, the rank in the absence of noise of a data matrix, is at a minimum. Once this minimum is found, the data is shifted linearly accordingly. Although this method works well for GC × GC data, we were interested in using a method that would be simple and effective for entire data runs so that the data could be compared and aligned as a whole. Also, we again wanted the capability of aligning the two dimensions using different techniques due to the different meanings of misalignment in each dimension.

Here, we present a method for the alignment of two dimensions using data resulting from the pyrolysis of Bacillus endospores. Pyrolysis is used to break the spores apart by high heat into volatile components that can be detected by DMS, as we have shown previously [37], [38]. We use a single file for reference and progress first in the Vc dimension, where the amount of shift from the unaligned file to the reference is determined, and the data then moved rigidly (without interpolation) by that amount. The amount of shift in the time dimension is then determined by matching features in the data, and the equation for a straight line is calculated between each successive matched feature. The subsequent piecewise-linear function is then applied to the data so that the features are aligned with those in the reference file. We will also provide evidence that our algorithm is able to compensate for slight changes in the gas flow based on the examination of the signal of the reactant ion peaks that result from the ionization of nitrogen, which can be considered a standard for the system.

Section snippets

Experimental

The data analysis was performed by code written in MATLAB (The Mathworks, Inc., Natick, MA) software Version 6.5.1.199709 Release 13. Several functions included with MATLAB were used. During the alignment of the Vc dimension, the interp and xcorr functions were used. For the time alignment, filtfilt (zero-phase digital filtering) and fir1 (windowed linear-phased finite impulse response filtering) were used. Three threshold values are required as user-input parameters. The first and second are

Results

A series of pyrolysis-differential mobility spectrometry experiments were conducted with a small volume of B. cereus spores suspended in water. The DMS data was collected throughout the pyrolysis event. These experiments were conducted under identical conditions with identical sample preparations and were repeated multiple times to examine the consistency of the data.

The data that results from the DMS is three-dimensional, with the two independent axes representing compensation voltage and scan

Discussion

Sensor data from repeat experimental runs can become misaligned due to varying experimental conditions, such as slight fluctuations in gas flow rate, varying moisture content between samples, instrument variability on different days, and temperature fluctuations. Although only a deactivated fused silica column was used in these experiments, if a separation column was used prior to DMS detection, the condition of the column could also cause variations across runs. We have demonstrated a method

Conclusions

The algorithm presented here can be applied to differential mobility spectrometry data to align in both the Vc and the time dimensions to a reference file. Alignment may be necessary due to slight experimental variations in heat, flow, or moisture. The alignment can aid in further data processing in which the files will be directly compared. This method can also be applied to data from other sensor modalities that have drift in two dimensions, in which the drift in one of the dimensions is

Acknowledgements

We would like to thank Dr. Abraham L. Sonenshein of Tufts University School of Medicine for providing the B. cereus spore samples. This project was partially sponsored by the Department of the Army, Cooperative Agreement DAMD-17-02-2-0006. The content of this paper does not necessarily reflect the position or the policy of the government, and no official endorsement should be inferred.

Melissa Krebs is a member of the technical staff at The Charles Stark Draper Laboratory. She received BS (2002) and MS (2003) in biomedical engineering from the University of Rochester in Rochester, NY. Her master's thesis work involved the study of protein interactions in the cellulosome of Clostridium thermocellum to investigate the mechanism for the efficient degradation of cellulose. She is currently investigating various application areas the differential mobility spectrometer (DMS) sensor

References (44)

  • R. Guevremont

    High-field asymmetric waveform ion mobility spectrometry: a new tool for mass spectrometry

    J. Chromatogr. A

    (2004)
  • B. Zitova et al.

    Image registration methods: a survey

    Image Vision Comput.

    (2003)
  • A.W. Toga et al.

    The role of image registration in brain mapping

    Image Vision Comput.

    (2001)
  • A. Sowmya et al.

    Modelling and representation issues in automated feature extraction from aerial and satellite images

    ISPRS J. Photogramm. Remote Sens.

    (2000)
  • E. Reiner et al.

    Characterization of normal human cells by pyrolysis gas chromatography mass spectrometry

    Biomed. Mass. Spectrom.

    (1979)
  • M.E. Parrish et al.

    Computer-enhanced high-resolution gas chromatography for the discriminative analysis of tobacco smoke

    Anal. Chem.

    (1981)
  • H.T. Mayfield, B.W., An algorithm for rapidly organizing gas chromatography data into data sets for chemometric...
  • J.A. Pino et al.

    Application of pyrolysis/gas chromatography/pattern recognition to the detection of cystic fibrosis heterozygotes

    Anal. Chem.

    (1985)
  • T.J. Bahowick et al.

    Correlation of quantitative analysis precision to retention time precision and chromatographic resolution for rapid short-column analysis

    Anal. Chem.

    (1995)
  • P.H.C. Eilers

    Parametric Time Warping

    Anal. Chem.

    (2004)
  • R.A. Miller, G.A. Eiceman, E.G. Nazarov, A.T. King, A MEMS radio-frequency ion mobility spectrometer for chemical agent...
  • E. Krylov et al.

    Field dependence of mobilities for gas-phase-protonated monomers and proton-bound dimers of ketones by planar field asymmetric waveform ion mobility spectrometer (PFAIMS)

    J. Phys. Chem.

    (2002)
  • Cited by (12)

    • Alpha function of FAIMS: Modeling uncertainty and experimental findings

      2022, Sensors and Actuators B: Chemical
      Citation Excerpt :

      However, few studies have investigated the theoretical derivation and experimental study of α2 and α4 because of the complex relationship among the CV, DV, and alpha function, as shown in Eqs. (10)–(12). Therefore, the reliability of α2 and α4 can only be evaluated using numerical analysis, and accuracy improvements are mainly based on the results of experiments, such as introducing several types of DV waveforms [20,39,40], applying a stronger dispersion voltage [41], adopting a more accurate method for drift tube fabrication [42], increasing the measurement time [43,44], and exploring new signal processing algorithm [45] and so on. However, these methods do not provide theoretical guidance or evaluation criteria.

    View all citing articles on Scopus

    Melissa Krebs is a member of the technical staff at The Charles Stark Draper Laboratory. She received BS (2002) and MS (2003) in biomedical engineering from the University of Rochester in Rochester, NY. Her master's thesis work involved the study of protein interactions in the cellulosome of Clostridium thermocellum to investigate the mechanism for the efficient degradation of cellulose. She is currently investigating various application areas the differential mobility spectrometer (DMS) sensor technology, including biowarfare detection and clinical diagnostics.

    Cristina Davis is an assistant professor of mechanical and aeronautical engineering at the University of California, Davis. She was formerly a principle member of the technical staff and group leader of bioengineering at The Charles Stark Draper Laboratory. Her research focuses on biological detection and analysis of biomarkers using the differential mobility spectrometer (DMS) sensor technology. Her previous research experience includes high-throughput electrophysiology systems, membrane biophysics, signal transduction pathways in cell volume regulation, voltage-sensitive fluorescent dyes, and microelectromechanical systems (MEMS) biosensor research. Dr. Davis earned her BS degree (1994) at Duke University with a double major in mathematics and biology. She went on to complete her MS (1996) and PhD (1999) in biomedical engineering at the University of Virginia focusing on ion channel electrophysiology and novel biosensor research. She then worked on silicon-chip based biosensors during a postdoctoral fellowship at the Johns Hopkins University, and she continued this work at Molecular Devices Corporation in Lausanne, Switzerland to develop a commercialized version of the technology. During her training, she was supported by an NIH Predoctoral Cardiovascular Training Grant (1996–1999) and a Cystic Fibrosis Foundation Postdoctoral Fellowship Award (2000–2001).

    View full text