DE102005061819B3 - Method for reducing a data record - Google Patents

Abstract

A computer method for the reduction of a data set consisting of N (n + 1) tuples (x <SUB> 1alpha </SUB>, x <SUB> 2alpha </SUB>, ..., x <SUB> nalpha </ SUB >, y <SUB> alpha </SUB>), where alpha = 1, 2, ..., N and n is at least 2, has the following steps: a) Creation of a first total data record by the y Values of the tuples are replaced by the sums of the y values of those tuples whose x <SUB> 1 </SUB> values lie in an interval around the x <SUB> 1 </SUB> value of the original tuple, where the intervals are chosen so that their widths correspond to the width of characteristic signals in the x <SUB> 1 </SUB> dimension, b) creating a second sum data set by using the y values of the tuples of the first sum data set obtained in a) are replaced by the sums of the y values of those tuples whose x <SUB> 2 </SUB> values lie in an interval around the x <SUB> 2 </SUB> value of the original tuple of the first sum data record, whereby the Intervals so ge are selected so that their widths correspond to the width of characteristic signals in the x <SUB> 2 </SUB> dimension, c) For all n> 2, if available: Create an nth sum data record by using the y values of the Tuples of the n-1 th sum data record obtained in b) are replaced by the y values of those tuples whose x <SUB> n </SUB> values are in an interval around the x <SUB> n </SUB> value of the original tuple of the n-1 th sum data set, the intervals being chosen so that their widths correspond to the width of characteristic signals in the x <SUB> n </SUB> dimension, d) Determine at least ...

Description

The The present invention relates generally to the field of reduction the dataset of records while largely avoiding the loss of relevant information becomes.

Especially The present invention relates to a process for the reduction of Records a computer program which, in conjunction with a computer, the inventive method executing, as well as a corresponding computer program on a computer readable Medium is stored.

The Size of records plays in their processing with the computer a crucial role. The increasing the size of records prolongs the Processing time and increased the demands on the system resources. For example larger storage media, improved memory management and increased processor performance required. Thus, large amounts of data often cause you to exponentially Increasing cost and also an increased amount of time and make so often the editing of this great records impossible with the average PC.

One Way to face these problems is it. the size of the accruing records to reduce.

One Area, which will be explained here by way of example only and on the Problems with the accumulating large amount of data often occur is the evaluation of records coupled chromatography-mass spectroscopy experiments.

in this connection is a sample to be examined first chromatographically in a Column separated, before the individual sample components are analyzed by mass spectroscopy become.

in the Case of LC-MS, d. H. Liquid Chromatography Mass Spectrometry (Liquid Chromatography - Mass Spectroscopy) experiments, becomes the mass spectroscopic part The analysis is often done with electrospray mass spectrometry. in this connection create records, which usually amounts to a few hundred megabytes to have more than a gigabyte.

These Size of the records is problematic both in terms of the evaluation and in the With regard to archiving.

In the prior art, for example, the following methods are known for reducing the size of LC-MS data sets:
On the one hand, it is possible to reduce the detection sensitivity of the mass spectrometric step. For the commonly used MCP (microchannel plate) detectors, this can be achieved by reducing the MCP voltage. With this measure, the scope of the data sets can be arbitrarily reduced. However, this reduction has the consequence that very weak signals are no longer detected or that the mass spectroscopic response factor - the ratio of the intensity of the mass spectral signal to the concentration of the analyte in the sample - decreases with decreasing analyte concentration and tends to zero. This behavior no longer ensures reliable quantification of small signals. This represents a decisive disadvantage, especially in comparative studies of recombinant therapeutic proteins, so-called comparability studies, since the aim of such studies is to be able to exclude the presence of unexpected signals with the greatest possible reliability.

Farther can be a so-called cut-off value in the data acquisition of the mass spectra be used. This procedure is the same a method that records with the record limited in scope become. The reduction comes from the fact that signals not recorded below the cutoff value. The disadvantage This method is that the correct cutoff value in the Difficult to determine in advance, and usually not the same Value for all individual spectra is suitable. The last point is therefore from Meaning, because the scope of nonspecific signals in particular by the presence of analyte ions of high concentration significantly is increased.

Finally, noise signals can be removed by statistical analysis of individual mass spectra. Such a method is described for example in the patent application WO 2005031791. Here, individual mass spectra are analyzed statistically with the aim of identifying non-specific, ie the noise attributable signals to remove these signals from the spectra. A similar procedure ren is also implemented in the under the brand name MassMap ^® (www.massmap.bregler.org) available software for the analysis of LC-MS data sets. In the case of MassMap ^® does not store the spectra from which the noise signals are removed done. The software provides a routine for calculating total ion current chromatograms that do not take into account the signals associated with noise. With the removal of peaks attributable to the noise signals and identified by statistical analyzes of individual spectra, a reduction in the amount of LC-MS data sets and an improvement in the signal-to-noise ratio of total ion current chromatograms are achieved , However, the consideration of the single spectra for signals at the detection limit means that also in this method, many signals that originate from analytes are erroneously assigned to the noise. As a result, a loss of sensitivity also results here.

adversely In all of these methods, it is thus that, although signals carrying the Analytes correspond, at the detection limit still reliably detectable are the ones mentioned Process these signals in single mass spectra as noise signals mis-identify.

In to lead the consequence These methods thus reduce the detectability of Signals at the detection limit.

The publication US 3,800,236 describes a circuit arrangement for compensating a drifting baseline (zero line) of an input signal to be measured which consists of successive peaks. The circuit includes: a peak detector including an integrator that integrates the signal variations over successive short measurement intervals and is reset to zero after each measurement interval, and also includes a comparator that compares the output of the integrator to a particular reference signal; provides a peak detection signal when the output of the integrator exceeds the reference signal during a measurement interval; a counter in which pulses can be counted at a deviation of the input signal from zero, a digital / analog converter, by which the readout result of the counter can be converted into an analog correction signal that is algebraically subtracted from the input signal to the zero line correction, and a Means for controlling the count operation in response to the peak detection signal of the peak detector such that in determining the presence of a peak in the input signal, the zero line compensation is stopped.

The publication US 5,995,989 A describes a method for compressing and filtering spectroscopy data. The method is for detecting peaks and filter data related to background noise, thereby reducing the data volume. The method monitors the value of each data point and compares it to previous data to determine if that data point is at or near a peak. The y values for each data point are continuously summed and averaged to determine the average background level. The deviation is determined for each subsequent data point and used to determine a threshold. Each subsequent data point is compared to the threshold, and if it is above the threshold, it is assumed to be part of or near a peak. At this time, averaging is paused until a subsequent data point is below the threshold. After acquiring all the peaks, data concerning the background noise may be omitted.

at the method according to the two last mentioned pamphlets, however, it is not possible, small Peaks that fall within the range below the threshold, so that they too are removed and thus essential Information lost through data reduction.

outgoing The object of the present invention was therefore to provide a method for the reduction of records available in which the detectability of signals at the detection limit as little as possible is reduced.

Especially It was an object of the present invention, a corresponding Method for reducing chromatography-mass spectrometry data sets Computer program that works in conjunction with a computer or a computer Computer system performs this method, and a corresponding computer program, which is stored on a computer readable medium available to put.

These The object is achieved by a method for the reduction of data sets as claimed 1 or 2, by a computer program according to claim 14, in connection with a computer performs the method according to the invention, as well as by a computer program according to claim 15, which is on a computer readable Medium is stored, solved.

Surprisingly the inventor of the present invention has found that if you take the statistical analysis of individual records the individual data acquisition processes, for example, using scans, sums of multiple individual data collections, the signals at the detection limit no longer falsely mimic the noise but are recognized as relevant signals.

Consequently the inventor came to the conclusion that with the method according to the invention ultimately reached a more accurate result, even though they sum Scans instead of individual scans analyzed statistically, and thus the absolute Signal values and the measurement errors summed.

Next these properties, the method of the invention differs also by some further aspects of for example in the patent application WO 2005031791 disclosed prior art.

in the Difference to the method of the prior art, the inventive method comes without an extra step to identify peaks. This allows a faster data processing and protects system resources.

The inventive method needed further in contrast to that disclosed in document WO 2005031791 Procedure has no prerequisites regarding the structure of the Noise while in WO 2005031791 e.g. in the case of Tof spectra a Poisson distribution Is accepted.

Finally, can the calculation of the mean and standard deviation as statistical Moments in the method according to the invention for example, only for Data points with non-zero intensity values occur.

Farther the inventor has surprisingly found that the selective detection of nonspecific peaks, such as noise peaks, is further enhanced when the individual scans are summed up so that the one to be added Orient scans to the width of a characteristic signal.

One such a characteristic signal is for example a measuring signal, that in the dimension in which it was measured, a particular one Part or a specific property of the measured object, for example a sample substance or an array to be observed more accurately describes.

• a) Erstellen eines ersten Summendatensatzes dadurch, dass die y-Werte der Tupel ersetzt werden durch die Summen der y-Werte derjenigen Tupel, deren x₁-Werte in einem Intervall um den x₁-Wert des ursprünglichen Tupels liegen, wobei die Intervalle so gewählt sind, dass ihre Breiten der Breite charakteristischer Signale in der x₁-Dimension entsprechen,
• b) Erstellen eines zweiten Summendatensatzes dadurch, dass die y-Werte der Tupel des in a) erhaltenen ersten Summendatensatzes ersetzt werden durch die Summen der y-Werte derjenigen Tupel, deren x₂-Werte in einem Intervall um den x₂-Wert des ursprünglichen Tupels des ersten Summendatensatzes liegen, wobei die Intervalle so gewählt sind, dass ihre Breiten der Breite charakteristischer Signale in der x₂-Dimension entsprechen,
• c) Für alle n > 2, falls vorhanden: Erstellen eines n-ten Summendatensatzes dadurch, dass die y-Werte der Tupel des in b) erhaltenen n-1-ten Summendatensatzes ersetzt werden durch die y-Werte derjenigen Tupel, deren x_n-Werte in einem Intervall um den x_n-Wert des ursprünglichen Tupels des n-1-ten Summendatensatzes liegen, wobei die Intervalle so gewählt sind, dass ihre Breiten der Breite charakteristischer Signale in der x_n-Dimension entsprechen,
• d) Ermitteln mindestens eines jeweiligen ersten statistischen Parameters des jeweiligen n-ten Summendatensatzes aus Schritt b) oder c),
• e) Ermitteln eines jeweiligen Schwellenwertes für die y-Werte aus dem jeweiligen ersten statistischen Parameter,
• f) Vergleichen der y-Werte des n-ten Summendatensatzes mit dem jeweiligen Schwellenwert,
• g) Erzeugen eines reduzierten Datensatzes durch Entfernen aller Tupel des zu reduzierenden Datensatzes, deren entsprechender y-Wert im n-ten Summendatensatz aus Schritt b) oder c) nicht größer als der Schwellenwert ist.

Thus, the present invention includes a method of reducing a data set consisting of N (n + 1) tuples (x _1α , x _2α , ... x _nα , _yα ), where α = 1, 2, ... , N and n is at least 2, the computer method comprising the following steps:

• a) Creating a first sum data set by replacing the y-values of the tuples with the sums of the y-values of those tuples whose x ₁ values lie at an interval around the x ₁ value of the original tuple are chosen so that their widths correspond to the width of characteristic signals in the x ₁ dimension,
• b) generating a second sum data set by replacing the y values of the tuples of the first sum data set obtained in a) by the sums of the y values of those tuples whose x ₂ values are in an interval around the x ₂ value of the original one Lie tuples of the first summation data set, wherein the intervals are chosen such that their widths correspond to the width of characteristic signals in the x ₂ dimension,
• c) For all n> 2, if available: Creation of an nth sum data set by replacing the y values of the tuples of the n-1th sum data set obtained in b) with the y values of those tuples whose x _n Values lie at an interval around the x _n value of the original tuple of the n-th summation dataset, the intervals being chosen such that their widths correspond to the width of characteristic signals in the x _n dimension,
• d) determining at least one respective first statistical parameter of the respective nth summation data record from step b) or c),
• e) determining a respective threshold value for the y values from the respective first statistical parameter,
• f) comparing the y values of the nth sum data record with the respective threshold value,
• g) generating a reduced data set by removing all tuples of the data set to be reduced whose corresponding y value in the nth sum data record from step b) or c) is not greater than the threshold value.

optional can also in step g) first all y-values of the record to be reduced are set to zero, their corresponding value in the nth summation dataset from step b) or c) not greater than that respective threshold is.

After that is then done by removing the tuples whose y values are in the previous one Step were set to zero, a reduced record created.

For example at time-dependent Image processing applications, such a process is useful and reduces the required storage volume without quality loss and improves additionally the picture quality through the elimination of noise signals.

Especially such a method is suitable, for example, if changes to be monitored by known objects over time. Conceivable is an application of the method thus under in the backup of camera-based surveillance systems.

at Chromatography-mass spectrometry analysis method becomes an additional Gain in selectivity with the method according to the invention achieved in that the evaluation of the signals of the summed Mass spectra are based on sums that span m / z ranges extend the expected width of mass spectrometric signals correspond.

in this connection can any chromatographic method, such as liquid chromatography, Capillary electrophoresis or gas chromatography with any suitable mass spectroscopic methods, e.g. Electrospray, Electron Impact Ionization, Collision Activation, or Neutralization Reionization MS be combined.

• a) für α ∊ {1, ...N} Summieren von einzelnen Massenspektren aus der Gesamtheit der Massenspektren einschließlich M_α, die innerhalb eines t_α einschließenden Zeitintervalls T_α aufgenommen wurden, zu jeweiligen ersten Summenmassenspektren M'_α bestehend aus n'_α Wertepaaren (x'_αk, y'_αk), wobei k ∊ {1,2, ..,n'_α} gilt, und die Breite der Zeitintervalle T_α in Abhängigkeit von der Breite von mindestens einem chromatographischen Peak im Zeitbereich um t_α bestimmt wird,
• b) für jedes x'_αk eines jeweiligen ersten Summenmassenspektrums M'_α Summieren der Werte y'_αm, deren entsprechende Werte x'_αm in einem m/z-Intervall X'_αk um x'_αk liegen, zu jeweiligen Summenwerten y''_αk, um ein jeweiliges zweites Summenmassenspektrum M''_α zu erhalten mit den Wertepaaren (x''_αk=x'_αk,y''_αk), wobei das m/z-Intervall X'_αk in Abhängigkeit von der Breite der entsprechenden massenspektrometrischen Peaks im Bereich um x'_αk bestimmt wird,
• c) Ermitteln mindestens eines jeweiligen ersten statistischen Parameters des jeweiligen zweiten Summenmassenspektrums M''_α,
• d) Ermitteln eines jeweiligen Schwellenwerts Y_α aus dem jeweiligen ersten statistischen Parameter,
• e) Vergleichen der Summenwerte y''_αk mit dem jeweiligen Schwellenwert Y_α,
• f) Erzeugen eines jeweiligen reduzierten Datensatzes durch Entfernen der Wertepaare (x_αj, y_αj) der jeweiligen Massenspektren M_α, für die die entsprechenden Summenwerte y''_αk nicht größer als der jeweilige Schwellenwert Y_α sind, wobei x''_αk = x_αj gilt.

Thus, one embodiment of the present invention is a method of reducing, for example, liquid chromatography mass spectrometry data sets by removing nonspecific signals from the data sets consisting of N mass spectra M _{α taken} during the generation of the data sets at successive times t _α and each n _α pairs of values (x _αj , y _αj ), where the x _{αj represent} the m / z values and the y _{αj represent} the intensity values and α ε {1, 2, ..., N} and j ε {1, 2, ..., n _α }, the method comprising the following steps:

• a) for α ε {1, ... N} summing individual mass spectra of the total of the mass spectra including M _α taken _α within a t _α enclosing the time interval T to respective first sum mass spectra M _'α consisting of n' _α Value pairs (x ' _αk , y' _αk ), where k _∈ {1,2, .., n ' _α }, and the width of the time intervals T _α as a function of the width of at least one chromatographic peak in the time domain by t _α it is determined
• b) for each x ' _{αk of} a respective first sum mass spectrum M' _α summing the values y ' _αm whose corresponding values x' _{αm lie} in a m / z interval X ' _αk by x' _αk to respective sum _values y " _αk to obtain a respective second sum mass spectrum M '' _α with the value pairs (x '' _αk = x ' _αk, y'' _αk ), wherein the m / z interval X' _αk depending on the width of the corresponding mass spectrometric peaks is determined in the range around x ' _αk ,
• c) determining at least one respective first statistical parameter of the respective second sum mass spectrum M " _α ,
• d) determining a respective threshold value Y _α from the respective first statistical parameter,
• e) comparing the sum _values y '' _αk with the respective threshold value Y _α ,
• f) generating a respective reduced data set by removing the value pairs (x _αj , y _αj ) of the respective mass spectra M _α , for which the corresponding sum values y " _{αk are} not greater than the respective threshold Y _α , where x" _αk = x _αj holds.

In this method, too, it is possible to subdivide step f) into two sub-steps, wherein in a first sub-step the y _αj values of the respective mass spectra M _α , for which the corresponding sum values y " _{αk are} not greater than the respective threshold value Y _α are set to zero. In a second sub-step, these value pairs are removed from the data set with the y _αj values set to zero.

According to a particularly preferred embodiment of the invention, a weighted summation takes place in step a), in which the weighting of the mass spectra M _β to be summed takes place with the magnitude | t _α -t _β | represents the temporal distance of the mass spectrum of the respective mass spectrum M _α monotonically decreasing function.

By decreasing monotonically, it is meant that the function value does not increase with increasing x becomes larger, that is, it follows from x ₁ <x ₂ that f (x ₁ ) ≥ f (x ₂ ).

Preferably, the function is strictly decreasing, that is, it follows from x ₁ <x ₂ that f (x ₁ )> f (x ₂ ).

In a further preferred embodiment of the invention, a weighted summation of the intensities takes place in step b), in which the weighting of the respective values y ' _αm one with the magnitude | x' _αm - x ' _αk | represents the distance of the respective value x ' _αm of x' _αk monotonically decreasing function.

Also Here again a strictly monotonically decreasing function is especially special prefers.

In a method according to the invention for the reduction of data sets for chromatography-mass spectrometric data sets, the respective time range T _α can furthermore be selected such that it corresponds essentially to the full width at half the height (FWHM) of the chromatographic peaks.

So The scan summation can optimally match an existing signal be adjusted. This ensures that the number the summed scans in a meaningful way out of the circumstances, and that only the scans are summed in one subarea, which makes sense a certain data similarity can be expected. In a special way, this step is the Sample separation in the upstream chromatography method Taken into account. If relatively wide peaks occur in the chromatogram, so is preferred only over These broad ranges are summed while, if the chromatography predominantly narrow peaks, even over these narrow areas will be summed up. This has the advantage that the mass spectra are summed in the areas that after the Chromatography step a largely correspond to purified sample component.

Also, in the second summation the summation range _αk X can be selected so that it is substantially the full width at half maximum (FWHM) corresponding to the mass spectrometric peak in the summation range. The peak width to be used for the interval size does not have to depend on the respective mass spectrum, but is generally a function of the m / z or x value. The peak width can be determined based on the resolution of the mass analyzer. It can either be largely independent of the m / z value or, as in the case of Tof analyzers, directly proportional to the m / z value, as in the case of ion traps and quadrupole analyzers. This selection of the summation range X _{αk avoids the accumulation of} too many unspecific signals originating from non-peak regions.

The Calculation of the statistical parameters of the second sum mass spectra is preferably done individually for every single spectrum. The common use of statistical Parameters for however, multiple spectra is also possible.

• da) Ermitteln eines ersten Schwellenwerts Y_α1 aus dem ersten statistischen Parameter,
• db) Ermitteln mindestens eines zweiten statistischen Parameters für diejenigen Wertepaare des zweiten Summenmassenspektrums M''_α, für die der Summenwert y''_αk größer als der erste Schwellenwert Y_α1 ist,
• dc) Ermitteln eines zweiten Schwellenwerts Y_α2 aus dem zweiten statistischen Parameter,
• dd) Wiederholen der Schritte da) bis dc) bis der Betrag der Differenz zwischen dem ersten Schwellenwert Y_α1 und dem zweiten Schwellenwert Y_α2 einen vorbestimmten Grenzwert ΔY nicht mehr übersteigt.

The determination of the threshold value Y _α in step d) comprises in a further embodiment the following steps:

• da) determining a first threshold Y _α1 from the first statistical parameter,
• db) determining at least one second statistical parameter for those value pairs of the second sum mass spectrum M '' _α , for which the sum value y '' _{αk is} greater than the first threshold value Y _α1 ,
• dc) determining a second threshold Y _α2 from the second statistical parameter,
• dd) repeating steps da) to dc) until the magnitude of the difference between the first threshold Y _α1 and the second threshold Y _α2 no longer exceeds a predetermined threshold ΔY.

Such A procedure allows the threshold to be set automatically for each Sum mass spectrum is calculated optimally adjusted and not must be determined individually by the skilled person.

Optionally, the steps c) to e) can be carried out for the entire value pair range of the respective second sum mass spectrum M " _α .

unterteilt, und die Schritte c) bis e) werden jeweils für einen Wertepaarteilbereich I_ακ durchgeführt.In one embodiment, in order to carry out steps c) to e), the respective second sum mass spectrum M " _{α is included} in value pair _subareas I _ακ

are divided, and the steps c) to e) are each performed for a _{Wertepaarteilbereich} I _ακ .

This advantageously makes it possible to adapt the range of value pairs in its extent to the corresponding peak width, so that, in accordance with a further embodiment, the respective value pair _portion I _{ακ is determined} as a function of the width of the mass spectrometric peaks when processing mass spectroscopic data sets.

wobei X(u)ακ die untere Grenze und X(o)ακ die obere Grenze des Wertepaarteilbereichs I_ακ sind, β einen Faktor größer als 5 darstellt, und δ für die Breite des Summierungsbereichs X'_ακ für m/z-Werte des Wertepaarbereichs I_ακ steht.In a specific embodiment, the value pairs _sections I _{ακ are determined} according to the following equations:

in which X (U) ακ the lower limit and X (O) ακ the upper limit of the value pair increment I are _ακ, β a factor greater than 5 is, and δ for the width of Summierungsbereichs X _'ακ for m / z values of the pair of values is the region I _ακ.

Furthermore, in the method according to the invention in step c) the mean value MW _α and the standard deviation SD _{α of} the non-zero intensity values can be used as statistical parameters and in step d) the threshold value Y _α according to the equation Y α = MW α + f (α) · SD α where f (α) represents a factor which preferably ranges from 0.5 to 5, and more preferably 2.

It However, it is clear to a person skilled in the art that he depends on the requirements to the accuracy of the data acquisition or to the size of the resulting records the factor f (α) each individually. Here, a small factor f (α) leads to a better data collection at the detection limit, but also too an elevated one Amount of data.

It It will be further understood by those skilled in the art that all in this invention disclosed embodiments as desired be combined with each other, without thereby by the scope of deviates from the present invention.

Detached from spectroscopic Analytical method is a method of reduction according to the invention of records For example, in the field of image processing conceivable.

One corresponding computer program that in conjunction with a computer directly or after execution a predetermined routine, a method according to any one of claims 2 to 13 executes, is also encompassed by the present invention.

Finally includes The present invention is a computer program based on a computer readable Media, such as floppy disks, hard drives, portable drives, CDs, DVDs, optical disks or tapes is stored.

Further Advantages and embodiments The present invention will become apparent from the embodiments as well as from the drawing.

It shows

1 a schematic overview of essential process steps of the method according to the invention;

2 a section of the total ion current chromatogram of a data set;

3 a section of the sum mass spectrum (sum of the scans from 13.85 min to 14.25 min) of the in 2 peaks marked A;

4 an enlarged section of the mass spectrum 3 in the environment of in 3 peaks denoted by a;

5 a comparison of sum mass spectra as obtained from differently processed data sets corresponding to the sum mass spectrum of 3 were obtained; and

6 a comparison of sum mass spectra as obtained from differently processed data sets corresponding to the sum mass spectrum of 3 were obtained.

The The following examples serve to illustrate the invention. However, they are not intended the scope of the present invention as defined by the claims is going to restrict.

Based on 1 the method according to the invention is explained schematically by way of example on the basis of a data set from an LC-MS experiment.

Such a data set consists here, for example, of a number of individual mass spectra, which are recorded in a regular chronological order. Such a single _{mass spectrum} M _α , which consists of a number of value pairs (x _αj , y _αj ), is shown schematically in the upper graph in FIG 1A shown.

These individual mass spectra are now summed at intervals in a first step. This is done by summing a certain number of spectra, which were taken in chronological succession. Here, the spectra summed in the summation interval T _α include the single mass spectrum M _α , which was recorded at the time t _α , and the spectra M _α-6 , M _α-5 , M _α-4 , M _α-3 , M _α-2 , M _α-1 , M _{α + 1} , M _{α + 2} , M _{α + 3} , M _{α + 4} , M _{α + 5} and M _{α + 6} .

The first sum mass spectrum M ' _α obtained in this way is shown in the lower graph in FIG 1A shown.

In a second step, shown in 1B , the thus obtained first sum mass spectrum M ' _{α are} summed up again. However, this summation occurs differently than in the first summation step within the respective first sum mass spectra. Here, the x _'αk values associated y' values are _{αk 'αk} each intermittently to a y' within an interval X _'αk value summed so that in the lower diagram in 1B shown second sum mass spectrum M '' _α arises.

From these second sum mass spectrum M _'α a first statistical parameters will now be determined, from which in turn a threshold value Y is _α determined.

This threshold is shown in the upper graph in 1C indicated by a horizontal line in the second sum mass spectrum. As can be seen, some y " _αk values in the second sum mass spectrum are below the threshold value Y _α .

These y " _αk values, which are smaller than Y _α , are assigned to an x" _αk value. This respective x '' _αk value in the second sum mass spectrum M '' _α corresponds to an x _αk value in a single _{mass spectrum} M _α . For the sake of clarity, this corresponding individual mass spectrum M _α is once again directly below the sum mass spectrum M " _α in the second graph from above in FIG 1C shown again.

In a last step, the pairs of values (x _αj , y _αj ) are removed from the corresponding respective single _{mass spectrum} M _α , whose corresponding y " _αk values are smaller than Y _α .

On this way, a reduced data set is created without the loss relevant information.

The Result of the data processing according to the invention is explained in more detail below with reference to an exemplary LCMS data set.

at the sample underlying the dataset is a Protein spot isolated from a polyacrylamide gel. The reduced and alkylated sample was tryptically digested.

The reversed phase analysis of the peptide mixture was carried out with the aid of a Vydac C ₁₈ capillary column (diameter 0.3 mm, length 150 mm) and a linear acetonitrile / water gradient (eluent A: water with 0.075 % (v / v) TFA; Eluent B: 80% acetonitrile / 20% water with 0.055% TFA; gradient: 5% B to 90% B in 30 minutes, flow rate 5 μl / min).

2 shows a section of the resulting total ion current chromatogram of the analyzed reduced, alkylated and tryptically digested protein poth sample. It can clearly be seen how the RP separation step separated the sample, so that the individual constituents of the sample to be analyzed had left the column in time sequence as eluate.

This eluate of the column was introduced directly into the source of an electrospray mass spectrometer. The device was a Q-Tof 2 from Micromass, GB. As the ion source, a nano-spray device was used. The time interval between the mass spectra was 2.15 s, the mass spectra were recorded from 200 amu to 2000 amu. Here, 1 (atomic mass unit) corresponds to 1.660538 × 10 ^-27 kg.

A sum spectrum of the thus recorded data record is in 3 shown. 3 shows a section of the sum mass spectrum (sum of the scans of 13.85 min to 14.25 min) of the in 2 with A marked peaks. The record has the properties given in Table 1: Table 1:

4 shows an enlarged section of the spectrum 3 , It is clearly visible that the individual peaks of the mass spectrum extend over a certain m / z range and have a FWHM value of about 0.15 amu.

This Record was using the method of the invention and comparatively processed with the method disclosed in WO 2005031791.

On the document WO 2005031791 is hereby incorporated by reference.

According to the method taught in document WO 2005031791, the data which correspond to a respective spectrum were first read out. Subsequently, statistical noise analysis was performed on these readout data to obtain statistical parameters related to the noise distribution. In the calculation of the expected values and the standard deviations as these statistical parameters, all data points were used in accordance with the description, and thus also those which had intensity values equal to zero. From the obtained statistical parameters, a limit was determined. This limit resulted from the expected value multiplied by twice the standard deviation. Finally, the relevant signals were in a respective spectrum by comparing the data points in the spectrum with the limit identified. Signals with an intensity below this limit were discarded.

In the processing of the same data set with the method according to the invention, the width of the time regions T _{α was set} constant to 7 data points, the time regions being distributed symmetrically around t _α . The width of the m / z intervals X ' _αk corresponded to x' _αj / 5000, these intervals also _being distributed symmetrically around x ' _αj .

Finally became the method with the factor for the determination of the Wertepaarteilbereiche β = 400 and with variable values for the factor f in the range of 2 to 2.5 for determining the threshold value carried out.

The file sizes that result from removing all data points from the original spectra whose associated intensities in the respective second sum mass spectra are equal to zero after performing steps a) to f) are summarized in Table 2 below compared with the resulting file sizes after the record processing according to WO 2005031791. Table 2:

to Evaluation of the resulting records were from the records using f = 2.15 and f = 2.50 calculates the sum mass spectra.

Comparisons of the corresponding sum mass spectra are given in FIGS 5 and 6 shown.

In 5 The "original smoothed" spectrum was obtained by subjecting the single spectrum a) derived from the original data set to binomial smoothing This smoothed spectrum is labeled b) The line drawn in this spectrum corresponds to the noise level of the spectrum, ie peaks are considered to be true above the line, while peaks which do not exceed the line are not necessarily to be regarded as significant The comparison spectrum c) was obtained after data processing with the method disclosed in WO 2005031791. Spectrum d) was measured with the invention Created a method for which a value of 2.15 was chosen.

Of the with the method according to the invention with f = 2.15 obtained record has according to the one shown above Table a circumference that is slightly smaller than the scope of the according to the WO 2005031791 received record. Despite about the same extent are in Mass spectrum, according to the method of the invention all those in spectrum b) marked with letters significant peaks clearly identifiable.

in the In contrast, in the mass spectrum produced according to WO 2005031791 the peaks b, e, i, r, t and w are no longer recognizable.

Analogous to 5 were the in 6 obtained with b) spectrum obtained by the fact that the derived from the original data set spectrum a) was subjected to a binomial smoothing.

The comparison spectrum c) was after the data processing with that in WO 2005031791 open barten process and is with the in 5 identical spectrum shown.

The d) was derived from a data set, the according to the inventive method with f = 2.50 was obtained.

It is obvious that even with the f = 2.50 accordingly obtained by the process according to the invention Spectrum the detectability of significant peaks with intensities at the Detection limit compared with that obtained according to WO 2005031791 Spectrum is greatly improved. This is true, though the corresponding Record has a scope that is only 55% of the scope of the according to WO 2005031791 record obtained.

There the computational effort to automatically evaluate LCMS datasets approximately square the record size increases, means halving the record size while improving Detectability of signals near the detection limit a substantial Advantage.

So can by the method according to the invention smaller records with very accurate information even with comparatively low-performance Computers swiftly to be edited.

The Nevertheless, accuracy at the detection limit allows a very high degree of accuracy detailed evaluation of the data.

Also the backup requires less system resources, as by the disregard of non-specific Information and through the restriction on the significant information the amount of data to be stored considerably is reduced.

Claims (15)

A method of reducing a data set consisting of N (n + 1) tuples (x _1α , x _2α , ... x _nα , y _α ), where α = 1, 2, ..., N and n at least 2, comprising the steps of: a) generating a first sum data set by replacing the y values of the tuples with the sums of the y values of those tuples whose x ₁ values are in an interval around the x ₁ value lie the original tuple, the intervals are chosen so that their widths correspond to the width of characteristic signals in the x ₁ dimension, b) creating a second sum data set in that the y values replace the tuples of the first summation data set obtained in a) are determined by the sums of the y-values of those tuples whose x ₂ values lie in an interval around the x ₂ value of the original tuple of the first sum data set, the intervals being chosen such that their widths are equal to the width of characteristic signals in the x ₂ dimension, c) For all n> 2, if available: Creation of an nth sum data set by replacing the y values of the tuples of the n-1th sum data set obtained in b) with the y values of those tuples whose x _n Values are at an interval around the x _n value of the original tuple of the n-th summation data set, the intervals being chosen such that their widths correspond to the width of characteristic signals in the x _n dimension; d) determining at least one e) determining a respective threshold value for the y values from the respective first statistical parameter, f) comparing the y values of the nth sum data record with the respective first statistical parameter of the respective nth summation data record from step b) or c) g) generating a reduced data set by removing all tuples of the data set to be reduced, whose corresponding y-value in the nth total data record from step b) or c) is not greater than that Threshold is. Method for reducing chromatography-mass spectroscopy datasets by removing nonspecific signals from data records which consist of a total of N mass spectra M _α , which are recorded at successive times t _α and which each have n _α value pairs (x _αj , y _αj ), where the x _{αj represent} the m / z values and the y _αj the intensity values and α ε {1,2, ..., N} and j ε {1,2, ..., n _α }, which is the comprising the following steps: a) for α ε {1, ... N} summing individual mass spectra of the total of the mass spectra including M _α taken _α within a t _α enclosing the time interval T, consist of respective first sum mass spectra M _'α from n ' _α value pairs (x' _αk , y ' _αk ), where k ε {1,2, .., n' _α }, and the width of the time intervals T _{α is determined} as a function of the width of at least one chromatographic peak in the time domain by t _α , b) for each x ' _{α k of} a respective first sum mass spectrum M' _α summing the values y ' _αm whose corresponding values x' _αm in a m / z interval X ' _αk by x' _αk , to respective sum _values y " _αk to obtain a respective second sum mass spectrum M" _α with the value pairs (x " _αk = x ' _αk , y '' _αk ), wherein the m / z interval X ' _{αk is determined} as a function of the width of the corresponding mass spectrometric peaks in the range around x' _αk , c) determining at least one respective first statistical parameter of the respective second sum mass spectrum M '' _α , d) determining a respective threshold value Y _α from the respective first statistical parameter, e) comparing the sum _values y " _αk with the respective threshold value Y _α , f) generating a respective reduced data set by Removing the value pairs (x _αj , y _αj ) of the respective mass spectra M _α , for which the corresponding sum values y '' _{αk are} not greater than the respective threshold Y _α , where x '' _αk = x _αj . Method according to claim 2, wherein in step a) a weighted summation takes place in which the weighting of the mass spectra M _β to be summed takes place with the magnitude | t _α -t _β | represents the temporal distance of the mass spectrum of the respective mass spectrum M _α monotonically decreasing function. Method according to claim 2, wherein in step b) a weighted summation takes place, in which the weighting of the respective values y ' _αm one with the magnitude | X' _αm -X ' _αk | represents the distance of the respective value x ' _αm of x' _αk monotonically decreasing function. Method according to one of claims 2 to 4, wherein the determination of the threshold value Y _α in step d) comprises the following steps: da) determination of a first threshold value Y _α1 from the first statistical parameter, db) determination of at least one second statistical parameter for those value pairs of the second sum mass spectrum M '' _α , for which the sum value y '' _{αk is} greater than the first threshold Y _α1 , dc) determining a second threshold Y _α2 from the second statistical parameter, dd) repeating the steps da) to dc) to the amount of the difference between the first threshold Y _α1 and the second threshold Y _α2 does not exceed a predetermined threshold ΔY. Method according to one of claims 2 to 5, wherein the steps c) to e) for the entire value pair range of the respective second sum mass spectrum M '' _{α are} performed. Method according to one of claims 2 to 6, wherein for carrying out steps c) to e) the respective second sum mass spectrum M '' _α in _{Wertepaarteilbereiche} I _ακ with

is divided, and the steps c) to e) in each case for a _{Wertepaarteilbereich} I _{ακ be} performed. The method according to claim 7, wherein the respective value pair _portion I _{ακ is determined} as a function of the width of the mass spectrometric peaks. Method according to one of claims 2 to 8, wherein the respective time range T _α substantially corresponds to the full width at half height, FWHM, of the chromatographic peaks. A method according to any one of claims 2 to 9 wherein the m / z interval X ' _{ακ is substantially equal to} the full width at half height, FWHM, of the mass spectrometric peaks. Method according to one of claims 7 to 10, wherein the value pair _subareas I _{ακ are determined} according to the following equations:

in which x (U) ακ the lower limit and x (O) ακ the upper limit of the value pair _portion I _ακ , β represents a factor which is greater than 5 and δ stands for the width of the m / z interval X ' _ακ in the value pair range I _ακ . Method according to one of claims 1 to 11, wherein in step c) the mean value MW _α and the standard deviation SD _{α of} the non-zero intensity values are determined as statistical parameters and in step d) the threshold value Y _{α is determined} according to the equation Y α = MW α + f (α) · SD α is determined, where f (α) is a factor. Method according to claim 12, where f (α) a value independent of α from 0.5 to 5, in particular 2. Computer program with program code, for carrying out the Method steps according to one of claims 1 to 13, when the program running in a computer becomes. Computer program with program code on one computer readable medium is stored to carry out the Method according to one of the claims 1 to 13 when running the program in a computer.