WO2009071241A1

WO2009071241A1 - Method for evaluating analyte-related signals from an ims chromatogram captured by means of ion mobility spectrometry

Info

Publication number: WO2009071241A1
Application number: PCT/EP2008/010100
Authority: WO
Inventors: Jörg Ingo BAUMBACH; Sabine Bader
Original assignee: Gesellschaft zur Förderung der Analytischen Wissenschaften e.V.
Priority date: 2007-12-03
Filing date: 2008-11-28
Publication date: 2009-06-11
Also published as: DE102007059461A1

Abstract

The invention relates to a method for evaluating analyte-related signals from an IMS chromatogram captured by means of ion mobility spectrometry, said chromatogram being created from a plurality of individual spectrums, a complete chromatogram or a part of a chromatogram, wherein the peaks showing up in the chromatogram are detected each by a gradual threshold value method, then approximated by an elliptic function and identified based on the detected elliptic parameters.

Description

"Method for Evaluating Analytical Signals from an IMS Chromatogram Recorded by Ion Mobility Spectrometry"

The invention relates to a method for evaluating analytical-related signals from an IMS chromatogram recorded by means of ion mobility spectrometry, which is formed from a series of individual spectra, a complete chromatogram or part of a chromatogram.

Ion mobility spectrometers, sometimes referred to as ion mobility spectrometers, are known and are finding ever wider applications, particularly since they have succeeded in miniaturizing such devices. Such ion mobility spectrometers usually have an ionization and reaction space which is separated from a drift space by an ion lattice. While in the ionization and reaction space the ions can be formed in different ways from a gas to be analyzed, which enters the ionization and reaction space through a gas inlet, the ions thus formed drift under the influence of the ion lattice along the drift space an electric field in the direction of a drift space end limiting Faraday plate, in which case an electrical signal is formed. A peculiarity of most ion mobility spectrometers is that these drift gas flows in the direction of the Faraday plate moving ions. Initially, this drift gas serves to prevent molecules or ions from adhering to the walls. The more important task, however, is to ensure that only ions on the lattice can reach the drift space and no neutral molecules of the analytes. Since only the ions formed from the analyte are electrically charged and thus can move in the electric field toward the Faraday plate, the neutral, i.e. the non-ionized molecules of the analytes are flushed out of the ionization and drift space.

In ion mobility spectrometry, three-dimensional data structures are often obtained by coupling with gas chromatographic columns, in particular multi-capillary columns, so-called IMS chromatograms. To describe the peaks as analyte-related signals from IMS chromatograms for individual Spectra are usually used Gaussian curves, so that each spectrum is processed individually and then closed from the two-dimensional to the three-dimensional case.

In particular, the sole peak detection based on the position on the drift time scale (often a so-called windowing technique with yes / no statement is used here) and / or the retention time leads to errors in the interpretation of IMS chromatograms. In addition, since mostly only (two-dimensional) calibration curves are considered, automatic spectral interpretations often fail even with ion mobility spectrometers coupled to a gas chromatographic column. Differentiation to IMS chromatograms stored in databases or to those of a previous measurement - for example, to take account of the influence of indoor air in exhaled air examinations - aggravates the problem.

Since the response of an ion mobility spectrometer under unknown external conditions, such as temperature, pressure or humidity variation as the main influencing parameter is usually not considered at all and if, then usually only for the two-dimensional case, the need arises, the general response function of a ion mobility spectrometer in three-dimensional space Considering the connection of disturbances analytically to describe.

The object of the invention is to provide a solution which enables automated recognition and unambiguous positioning of a peak in an IMS chromatogram, both with regard to the position with respect to the drift time and the retention time in order to identify an analyte as accurately as possible.

This object is achieved in a method of the type described according to the invention in that the peaks appearing in the chromatogram each detected by a stepwise threshold method, subsequently approximated by an ellipse function and identified by the determined ellipse parameters. Surprisingly, it has been found that a peak in an IMS chromatogram can be described analytically by means of various parameters in three-dimensional space by setting the parameters of an ellipse.

It is preferably provided that the measured data of the IMS chromatogram are first subjected to preprocessing in order to minimize measurement data distortions due to environmental and / or device influences of the ion mobility spectrometer. Such preprocessing, e.g. For suppressing the noise or the like., Can be done in different ways, as exemplified below.

It is further provided that the local minimum in the histogram of the intensity values is used for the determination of the endpoint of the stepwise threshold method. It is provided that the sequence of the intervals underlying the stepwise threshold method is set so that the number of points within the intervals increases constantly.

In addition, it is envisaged that in the stepwise procedure of the threshold method at each stage, first all data points that fall in the interval belonging to the respective stage are considered as peak points and these are then processed so that it can be determined which of these peak points belong to which peak, and from this gradually Peaklisten be created.

Furthermore, it is provided that the stepwise peak lists are linked to one another in such a way that also peaks that are not baseline-separated can be detected separately by adding entries of the current peak list in the event that the associated peak region contains more than one peak position of the previous stage. be replaced by the entries in the previous level peak list.

In a further advantageous embodiment, it is provided that the ellipse parameters determined for the respective peak are stored in a database, which is stored in a database. - A -

the analytes characterized characterizing data sets are identified and the respective peak is identified by agreement with a data set.

Since the center, the height and both semiaxes are fixed in an ellipse, it is possible to use the parameters of the ellipse function determined for the respective peak to describe physical data such as temperature, pressure or humidity, so that a peak as such is clearly recognizable , It is thus analytically decidable whether a peak is present if a signal has a certain shape at a certain point. It can also be distinguished by the stepwise approach of peak detection, whether there may be two peaks that overlap in space.

The invention is explained in more detail below by way of example with reference to the drawing. This shows in

Fig. 1 is a typical spectrum of an ion mobility spectrometer;

2 shows a heat map of an IMS chromatogram,

3 the representation of a baseline correction,

4 shows the spectra of a chromatogram in side view with tailing in the area to the right of the reaction ion peak,

5 shows a median spectrum and the determined log normally distributed function (top) and the difference (bottom),

6 a heat map from the raw data or measured data (left) and a heat map with a tailing function,

7 further heatmaps, where on the left data before the transformation, in the middle data after wavelet-shooting and on the right after wavelet-denoising dar- are put,

8 shows the illustration according to FIG. 7 in a side view of the spectral series, FIG.

9 shows a histogram of the intensity values of the original measured data (left) and at the end of the preprocessing of the measured data (right),

10 shows a representation of the determined threshold values,

11 a typical peak (left), an ellipse with the parameter a (middle) and an ellipse with the parameter b (right),

12 shows a heat map of the original measured values (left), a heat map at a higher threshold (center) and on the right a heat map at a lower threshold,

13 is an enlarged view of FIG. 12, in which different peaks are identified,

14 shows the representation of an ellipse adaptation of a single peak with preprocessed data (left) with the original measurement data (center) and with data transformed and denoised from the axis (right) and in FIG

15 shows a representation of original data (left) and ellipses (right) obtained by means of the method according to the invention.

In a known ion mobility spectrometer, not shown in the drawing, the analytes of a sample to be examined first enter the ionization region of the ion mobility spectrometer. There they are ionized in different ways and the molecules that are now charged can enter the drift space of the ion mobility spectrometer. At the Faraday plate of the ion mobility spectrometer a signal is tapped. Often done a pre-separation of the analytes by means of gas chromatographic methods, for example by means of multi-capillary columns. Then usually a whole series of spectra is recorded. This is called an IMS chromatogram.

The resulting data of a spectrum can be described as follows:

S = (Zi,.., Z2θOθ)

where z, are the values of the intensity of the ions of the analytes reaching the Faraday plate at a given drift time x, (i = 1, ..., 2000), the range of values for i being arbitrarily assumed once from 1 to 2000 , Usually, the drift times x, are kept equidistant. In Fig. 1, such a spectrum is shown.

A special feature is that the so-called reaction ion peak is always present. This, in effect, represents the reservoir for e.g. via ion-molecule reactions the analytes can be indexed. However, the information relevant for the measurements is usually available as comparatively small peaks. It is usually difficult to detect these peaks in the noise. In addition, they are often very close to each other.

In order to achieve a better resolution, it can be achieved with the aid of chromatographic columns that the analytes successively reach the ionization space of an ion mobility spectrometer. Thus, as an additional dimension, the

Retention time introduced: y ,, j = 1 500. Again, the range of values for j is 1 to

500 chosen arbitrarily.

This gives the matrix of intensity values z _u

S - (Si, ...., S5θθ) - (Z | j) ι = 1,, 2000, j = 1,, 500-

If you put all the spectra one behind the other, you get a so-called heat map, ie, a quasi-top view, as shown in Fig. 2. For example, such an IMS chromatogram can be from a single breath measurement. This results in two major problems. The first is the high dimensionality and redundancy, since a single such measurement results in millions of data points, most of which are in noise, and a number of comparatively significantly fewer signals from analytes. The second problem is that the data varies in all three dimensions, ie not only in the peak height, but also in the position of a peak in terms of variable drift and retention time.

Therefore, not only the question arises as to what the essential information within this data is, but also what are the variables that are likely to compare such measurements.

For example, because the peaks are correlated with analytes in the exhaled air, they represent the relevant information and are in some ways the answer to both questions asked. Therefore, it is necessary to find a way to characterize such peaks while reducing the number of data while losing little informational content.

By way of example, it should be assumed that a data record to be considered should consist of 500 spectra of a measurement, each containing 2000 data points distributed over a drift time of 50 ms.

First, data on drift time to account for factors such as ambient pressure and temperature are converted into the so-called reduced mobility and the retention time is corrected for the comparability of the individual measurements in view of the varying temperatures in the chromatographic column.

It is preferable to carry out the following steps first: 1. Baseline correction

2. Correction of the influence of device and environmental parameters on the drift time

3. Correction of the temperature influence on the retention time

In Fig. 3, the effect of the baseline correction is illustrated by, if necessary, setting the baseline and the associated noise to a value around zero.

By using the value 1 / K ₀ instead of the drift time itself as the x-axis, the influence of device parameters such as electric field strength E or drift path L and environmental parameters such as pressure p or temperature T can be taken into account:

K ₀ ^% 1000 L p T ₀

The temperature effect on the retention time can preferably be effected on the account of temperature variations by normalizing to a specific value, here 30 ⁰ C,:

^y f _Vi (1 + fcr AT) + y] AT k _q

In addition, there are methods that allow to consider the so-called reaction ion peak (RΙP) tailing. This tailing is shown in FIG.

It turns out that the peak at x-value 0.5 takes some time to return to the baseline. By the introduction of a log-normally distributed Function can be considered this, due to the individual device properties influence. FIG. 5 shows a median spectrum and the determined log-normal distributed function (top) and the difference (bottom). It can be clearly seen that other peaks lying in the region of the reaction ion peak are more prominent.

The basic structure of the function taking this into consideration is reproduced below.

θ = x ^s - ^m exp (σ ² )

A penalty has been introduced which allows to optimize the adaptation of the functions. It can be seen from FIG. 6 that, starting from the raw data (left), by means of a function considering this tailing, above all the signals nearer the reaction ion peak emerge more clearly (right) - this is especially evident in the range between 1.97 and 1.18.

Known methods such as wavelet transformation can also contribute to smooting and denoising. In FIGS. 7 and 8, the data before the transformation are shown on the left, in the middle those after wavelet-shooting and on the right those after wavelet-denoising; in Fig. 7 in associated heat maps and in Fig. 8 in side elevation of the spectral series. It is also clear from this that in this way the relevant signals emerge more clearly from the noise.

All of these aforementioned preprocessing steps are known. However, they do not solve the task of identifying the relevant peaks, but only make them more visible visually. For this purpose, the threshold values of the intensity ranges are first determined, which subsequently allow a stepwise procedure. For this, the histograms of the intensities are created.

In Fig. 9, the histogram of the original data is shown (left) and that obtained at the end of the above preprocessing procedure (right).

Surprisingly, it turns out that there is a local minimum that can be used to establish a threshold.

The threshold values, as shown in FIG. 10, are now established by first selecting a comparatively large range with respect to the signal intensity between the intensity maximum and half of this value - see FIG. 10, then another one down to around 1.5 , the next to around 0.9, and so on, creating an ever-narrower sequence whose final product is the threshold defined by the histogram.

The sequence of the intervals I _k in the range between the associated value i _k and the maximum of the matrix Zy ₁ is determined so that each I _{k is contained} in l _{k +} i:

I _k = [i _fc , max ^)] C [i _{k +} i, max (2 _y )] = I ₊ i V fc e [1,. , , , ny,

Here, n _{s is} the number of intervals, for example set to the value 25.

It has been found that a selection of the sequence such that the number of points in the intervals increases constantly is more suitable than a selection of the sequences with constantly increasing interval boundaries. Thus, it can be seen from FIG. 10 that the first interval has been selected such that only the reaction ion peak is found there. In the lower ranges of intensities where many peaks occur, the data is sampled more accurately. By defining the intervals in this way, the peaks can be identified in a stepwise process by the following route, each interval being associated with a stage of the process.

The first step at each stage of the procedure is to place the points of the matrix in the "peak" areas, where

applies, and "no peak" where

applies, can be split. This results in a matrix containing only the values 1 and 0.

In the next step, the peaks are separated from each other. The procedure is as follows: Starting with a matrix of independent data points. Spectrumwise segments of equal values (0 or 1) are now noted. If adjacent segments consist of the same values, they are grouped together into regions. In this way, different peak areas are differentiated from each other, so that it is then known which points of the data matrix belong to which peak.

After the detection of the peaks, they should now be characterized. It has surprisingly been found that it is possible to characterize a peak region with as few parameters as possible when using ellipsoids.

Usually applies to an ellipse {x - XQ) _ ² (y - y ₀ ) ² _{= i} α 2 Iβ

where a, b are the two semiaxes and (xo.yo) is the midpoint, the axes being chosen to be parallel to the coordinate axes, drift time and retention time.

First the peak position (xo.yo) is determined. The extent is determined by the parameters a and b. The corresponding specification of the peak action for a peak A is defined as follows:

y * = Vr »3 = ^m i ^aχ { ^z ij \ ^z ϋ ^ ^A l

A = {zi _j

belongs to peak A}

In Fig. 11, a typical peak is shown on the left. In the middle figure, an ellipse is drawn with the parameter a, in the middle with the parameter b.

In this case applies

α = miii (rc ₀ - αi, α ₂ - .τo), α, χ = Xj ', i' = min (i \ 3 eg _tj 6 A) α ₂ = X {", i" = max (« | 3% € A) or b = yo - Vr, / = min {j | = ^ € A)

This procedure results in a peak list P _k for each step of the stepwise process. Such a typical list with the values for x ₀ , yo, and the ellipse parameters a and b is shown below:

In addition to the parameters mentioned, the maximum height is indicated.

In order to be able to detect overlapping peaks as well, the procedure is as follows. It can be seen from Figure 12 that simply setting a threshold to distinguish between noise and peak either results in the two peaks being considered as one peak at low retention times, or the other two peaks being lost altogether walk. The original values are shown on the left, while in the middle the heatmap is shown at a higher threshold, which allows the two lower peaks with retention times greater than 20 to find. In this case, however, the larger peak is not resolved at lower retention times.

In other words, while the four signals A to D are plotted on the left of Fig. 13, for the largest threshold value at which the peaks A and B are discriminated, the peaks C and D are not resolved. On the other hand, in the case where the peaks C and D are separated, the peaks A and B disappear.

Therefore, the peak list is compared with the previous step PM for each step Pk. If one of the new peak regions then contains more than one peak of the previous key, this is an indicator of a multiple peak. Thus, the solution is to maintain the signals of step (k-1) for this region in step k and to discard the newly found, larger peak region containing multiple old peaks.

Finally, the area F of the ellipse is determined by the following equation:

F = π a b.

The last step is to optimize the height characterization of the peaks. For this purpose, the completely preprocessed data, which are shown on the left in FIG. 14 for the ellipse adaptation of a single peak, are no longer used, since these differ significantly in their height from the original data, which is represented in the center of FIG. 14 after introduced axis transformations are. Instead, to determine the peak height, the only axis-transformed and denoised data are used, which are shown in the right-hand image of FIG. 14.

In summary, the original data are shown in FIG. 15 on the left and the ellipses obtained by means of the method according to the invention on the right. For simplification the ellipses are marked with the numbers 1 to 9 on the right and also marked in the left rows of figures.

Claims

claims:

1. A method for evaluating analytical-related signals from an IMS chromatogram recorded by means of ion mobility spectrometry, which is formed from a series of individual spectra, a complete chromatogram or part of a chromatogram, characterized in that the peaks appearing in the chromatogram are each determined by a stepwise Threshold method detected, subsequently approximated by an ellipse function and identified on the basis of the determined ellipse parameters.

2. The method according to claim 1, characterized in that the measured data of the IMS chromatogram are first subjected to preprocessing in order to minimize measurement data distortions due to environmental and / or device influences of the ion mobility spectrometer.

Method according to claim 1 or 2, characterized in that the local minimum in the histogram of the intensity values is used for the determination of the end point of the stepwise threshold method.

Method according to claim 3, characterized in that the sequence of the intervals underlying the stepwise threshold method is determined so that the number of points within the intervals increases constantly.

5. The method according to claim 4, characterized in that in the stepwise procedure of the threshold value method at each stage, first all data points that fall in the interval belonging to the respective stage, are regarded as peak points and these are then processed so that it can be determined which of these Peak points to which peak belong, and from this gradually peaklists are created.

6. The method according to claim 5, characterized in that the stepwise peak lists are linked to one another such that also peaks that are not baseline-separated can be detected separately by entries in the current peak list in the event that the associated peak region has more than one peak position previous stage is replaced by the entries in question the peak list of the previous stage.

7. The method according to one or more of claims 1 to 6, characterized in that the determined for the respective peak ellipse parameters are compared with stored in a database, different analytes characterizing records and the respective peak is identified by agreement with a record.

8. The method according to one or more of claims 1 to 7, characterized in that the parameters determined for the respective peak of the ellipse function for the description of physical data, such as temperature, pressure or humidity are used.