COMPUTER-IMPLEMENTED METHOD FOR IDENTIFYING AT LEAST ONE PEAK IN A MASS SPECTROMETRY RESPONSE CURVE - Patent application

Abstract

A computer-implemented method (110) for identifying at least one peak in a mass spectrometry response curve is proposed, comprising the steps of: a) providing (112) at least one mass spectrometry response curve by using at least one mass analyzer (114), and b) identifying (116) a start point and an end point of at least one peak in the mass spectrometry response curve by evaluating the mass spectrometry response curve by using at least one trained model, the model being trained using a deep learning regression architecture (118).

Description

[Technical field]

[0001] The present invention relates to a computer-implemented method for identifying at least one peak in a mass spectrometry response curve and an apparatus for monitoring at least one analyte in a sample. The proposed method and apparatus can be used in the technical field of mass spectrometry, in particular in liquid chromatography-mass spectrometry. [Background technology]

[0002] Current liquid chromatography-mass spectrometry (LC-MS) data processing typically requires manual data review of all acquired data and then manual correction of approximately 5-20% of the results due to high error rates, performed by trained LC-MS operators through tedious visual analysis of hundreds of chromatographic plots.

[0003] Furthermore, automated methods are known for identifying and characterizing spectral peaks in a spectrum.

[0004] WO 2012 / 047417 and US 8,428,889 describe a method for automatically identifying and characterizing spectral peaks in a spectrum generated by an analytical instrument, the method comprising the steps of receiving a spectrum generated by the analytical instrument, automatically subtracting a baseline from the spectrum to generate a baseline-corrected spectrum, automatically detecting and characterizing spectral peaks in the baseline-corrected spectrum, and reporting the detected and characterized spectral peaks to a user. A list of adjustments to be made to the detection and characterization step from the user is received. Exit values ​​used in the detection and characterization step are adjusted based on the list of adjustments. The automatic detection and characterization of spectral peaks is repeated in the same spectrum or in one or more different spectra.

[0005] US Patent No. 7,219,038 describes a protocol for analysis and peak identification in spectroscopic data. A Bayesian method is used to automatically identify peaks in a data set. After identifying peak shapes, the method tests the hypothesis that a given number of peaks will be found within any given data window. If a peak is identified within a given window, a likelihood function is maximized to estimate the peak location and amplitude.

[0006] U.S. Patent No. 7,720,612 describes a method for resolving convoluted peaks in a chromatogram into one or more component peaks using peak resolution values. The peak method of this invention determines empirical peak resolution values ​​for "well-defined" or "separate" peaks in the data, and then extrapolates these empirical resolution values ​​to peaks in neighboring regions to predict the number of component peaks at a given peak position. The predicted peak resolution values ​​are compared to the observed peak resolution values ​​of low-resolution or convoluted peaks to determine the number of component peaks in the convoluted peak.

[0007] International Publication No. 2019 / 092836 describes that a model constructed by performing deep learning using accurate peak information and images obtained by imaging a large number of chromatograms as learning data is stored in advance in a trained model storage unit. When chromatogram data for a target sample obtained by an LC measurement unit is input, an image generation unit captures a chromatogram and generates an input image in which one of two areas on both sides of a chromatogram curve in the obtained image is filled. A peak position estimation unit inputs pixel values ​​of the input image to a neural network using a trained model, and obtains position information about the start and end points of the peak as well as peak detection accuracy as output. A peak determination unit determines the start and end points of the peak based on the peak detection accuracy.

[0008] European Patent Application Publication No. 3 467 493 describes that, prior to measurement, an analyst selects a chromatogram (a chromatogram covering a specific wavelength or all wavelengths) for peak detection and specifies a judgment value for the number of peaks via an input unit as a measurement termination condition. When the measurement unit executes the measurement, a chromatogram generation unit generates a chromatogram based on collected data in almost real time, and a peak detection unit detects peaks on the chromatogram. A measurement termination condition judgment unit counts the number of detected peaks and judges that the measurement termination condition is met when the counted number reaches the judgment value for the number of peaks, and a measurement termination timing judgment unit instructs the analysis control unit to terminate the measurement when a predetermined time has passed since the judgment.

[0009] Risum Anne Bech et al., “Using deep learning to evaluate peaks in chromatographic data”, TALANTA, ELSEVIER, Amsterdam, NL, vol. 204, 22 May 2019, pages 255-260, XP085747637, ISSN: 0039-9140, DOI: 10.1016 / J.TALANTA.2019.05.053, describe that the analysis of untargeted gas-chromatographic data is time-consuming and still involves several manual steps in the analysis that require data analysis expertise. One of these is the need to define whether each resolved component represents a peak suitable for integration or not. As peaks can vary both in shape and in their position on the elution time axis, this presents a problem that cannot be easily solved by the application of linear classifiers such as PLS-DA (Partial Least Squares Regression for Discriminant Analysis). A convolutional neural network classifier is described to handle these shifts and changes in shape.

[0010] Despite the advantages of the known methods and devices, these known methods and devices only provide information on whether a peak is present or not, but do not guarantee a reliable determination of the area under the peak.

[0011] It is therefore desirable to provide a computer-implemented method for identifying at least one peak in a mass spectrometry response curve and an apparatus for monitoring at least one analyte in a sample that addresses the above-mentioned technical problems. In particular, a method for identifying at least one peak in a mass spectrometry response curve and an apparatus for monitoring at least one analyte in a sample that allows for a reliable and automatic determination of the peak area of ​​the analyte peak in a chromatogram should be provided. Summary of the Invention

[0012] This problem is addressed by a computer-implemented method for identifying at least one peak in a mass spectrometry response curve and an apparatus for monitoring at least one analyte in a sample, the method and apparatus having the features of the independent claims. Advantageous embodiments which may be realized in either a single aspect or in any combination are set out in the dependent claims.

[0013] When used below, the terms "having", "comprises" or "includes" or any grammatical variants thereof are used in a non-exclusive manner. These terms can therefore refer both to the situation in which no further features are present in the entity described in this context besides the features introduced by these terms, and to the situation in which one or more further features are present. As an example, the expressions "A has B", "A comprises B" and "A includes B" can refer both to the situation in which no other elements are present in A other than B (i.e. the situation in which A is exclusively composed of B) and to the situation in which one or more further elements are present in entity A other than B, such as element C, elements C and D, or even further elements.

[0014] Furthermore, it should be noted that the terms "at least one" or "one or more," or similar expressions, indicating that a feature or element may be present one or more times, are typically used only once when introducing each feature or element. In the following, in most cases, when referring to each feature or element, the expressions "at least one" or "one or more" will not be repeated, despite the fact that each feature or element may be present one or more times.

[0015] Furthermore, when used below, the terms "preferably", "more preferably", "particularly", "more particularly", "particularly" or "more particularly" or similar terms are used in relation to any feature without limiting the possibility of substitution. Thus, the features introduced by these terms are optional features and are not intended to limit the scope of the claims in any way. The present invention may be implemented by using alternative features, as would be understood by a person skilled in the art. Similarly, features introduced by "in an embodiment of the invention" or similar expressions are intended to be optional features without any limitation on alternative embodiments of the invention, without any limitation on the scope of the invention, and without any limitation on the possibility of combining the feature introduced in such a way with other optional or non-optional features of the invention.

[0016] In a first aspect of the present invention, a computer-implemented method for identifying at least one peak in a mass spectrometry response curve is disclosed.

[0017] As used herein, the term "computer-implemented method" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, a method involving at least one computer and / or at least one computer network. The computer and / or computer network may comprise at least one processor configured to perform at least one of the method steps of the method according to the invention. Preferably, each of the method steps is performed by the computer and / or computer network. The method may be performed fully automatically, specifically without the need for user interaction. As used herein, the term "automatically" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, a process performed fully by at least one computer and / or computer network and / or machine, specifically without the need for manual action and / or user interaction.

[0018] As used herein, the term "mass analysis" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning. The term may specifically refer to, but is not limited to, an analytical technique for determining the mass-to-charge ratio of ions. Mass analysis may be performed using at least one mass analyzer. As used herein, the term "mass analyzer", also referred to as "mass analyzer", is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning. The term may specifically refer to, but is not limited to, an analyzer configured to detect at least one analyte based on its mass-to-charge ratio. The mass analyzer may be or may comprise at least one quadrupole analyzer. As used herein, the term "quadrupole mass analyzer" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning. The term may specifically refer to, but is not limited to, a mass analyzer that includes at least one quadrupole as a mass filter. A quadrupole mass analyzer may include multiple quadrupoles. For example, a quadrupole mass analyzer may be a triple quadrupole mass spectrometer. As used herein, the term "mass filter" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, a device configured to select ions injected into the mass filter according to their mass-to-charge ratio m / z. The mass filter may include two pairs of electrodes. The electrodes may be rod-shaped, and in particular cylindrical. In the ideal case, the electrodes may be hyperbolic. The electrodes may be identically designed. The electrodes may be arranged to extend parallel along a common axis, for example the z-axis.The quadrupole mass analyzer may include at least one power supply circuit configured to apply at least one direct current (DC) voltage and at least one alternating current (AC) voltage between two pairs of electrodes of the mass filter. The power supply circuit may be configured to hold each opposing electrode pair at the same potential. The power supply circuit may be configured to periodically change the sign of the charge of the electrode pairs such that stable trajectories are possible only for ions within a certain range of mass-to-charge ratios m / z. The trajectories of the ions in the mass filter may be described by the Mathieu differential equation. To measure ions of different m / z values, the DC and AC voltages may be varied over time such that ions having different m / z values ​​may be transmitted to the detector of the mass analyzer.

[0019] The mass spectrometer may further comprise at least one ionization source. As used herein, the term "ionization source", also referred to as "ion source", is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. This term may specifically refer to, but is not limited to, a device configured to generate ions from, for example, neutral gas molecules. The ionization source may be or may comprise at least one ionization source selected from the group consisting of at least one gas phase ionization source, such as at least one electron impact (EI) source or at least one chemical ionization (CI) source, at least one plasma desorption (PDMS) source, at least one fast atom bombardment (FAB) source, at least one desorption ionization source, such as at least one secondary ion mass spectrometry (SIMS) source, at least one laser desorption (LDMS) source, and at least one matrix assisted laser desorption (MALDI) source, at least one spray ionization source, such as at least one thermospray (TSP) source, at least one atmospheric pressure chemical ionization (APCI) source, at least one electrospray (ESI) source, and at least one atmospheric pressure ionization (API) source.

[0020] The mass analyzer may include at least one detector. As used herein, the term "detector" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, a device configured to detect incoming ions. The detector may be configured to detect charged particles. The detector may be or include at least one electron multiplier.

[0021] The mass analyzer, in particular the detector and / or at least one evaluation device of the mass analyzer, can be configured to determine at least one mass spectrum of the detected ions. As used herein, the term "mass spectrum" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term can specifically, but not exclusively, refer to a two-dimensional representation of signal intensity versus charge-to-mass ratio m / z, where the signal intensity corresponds to the abundance of the respective ion. The mass spectrum can be an image made up of pixels. The at least one evaluation device can identify analytes in the sample that can integrate the signal detected by the detector within a certain m / z range to determine the intensity of the pixels of the resulting mass spectrum. In particular, the evaluation device can be configured to correlate a known mass to the identified mass or can be configured by a characteristic fragmentation pattern.

[0022] The mass analyzer may be or comprise a liquid chromatography mass analyzer. The mass analyzer may be connected to and / or comprise at least one liquid chromatograph. The liquid chromatograph may be used for sample preparation for the mass analyzer. Other embodiments of sample preparation may be possible, such as at least one gas chromatograph. As used herein, the term "liquid chromatography mass analyzer" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, a combination of liquid chromatography and mass spectrometry. The mass analyzer may comprise at least one liquid chromatograph. The liquid chromatography mass analyzer may be or comprise at least one high performance liquid chromatography (HPLC) device or at least one micro liquid chromatography (μLC) device. The liquid chromatography mass spectrometry device may comprise a liquid chromatography (LC) device and a mass spectrometry (MS) device, in this case a mass filter, where the LC device and the mass filter are coupled via at least one interface. The interface coupling the LC device and the MS device may comprise an ionization source configured to generate molecular ions and transfer the molecular ions to a gas phase. The interface may further comprise at least one ion mobility module disposed between the ionization source and the mass filter. For example, the ion mobility module may be a high field asymmetric waveform ion mobility spectrometry (FAIMS) module.

[0023] As used herein, the term "liquid chromatography (LC) device" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, an analytical module configured to separate one or more analytes of interest of a sample from other components of the sample for detection of the one or more analytes by a mass spectrometer. The LC device may include at least one LC column. For example, the LC device may be a single-column LC device or a multi-column LC device having multiple LC columns. The LC column may have a stationary phase through which a mobile phase is pumped for separation and / or elution and / or transport of the analytes of interest. The liquid chromatography mass spectrometer may further include a sample preparation station for automated pre-treatment and preparation of samples, each of which includes at least one analyte of interest.

[0024] As used herein, the term "sample" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning. The term can specifically refer to any test sample, such as, but not limited to, a biological sample and / or an internal standard sample. The sample can contain one or more analytes of interest. For example, the test sample may be selected from the group consisting of physiological fluids including blood, serum, plasma, saliva, ocular lens fluid, cerebrospinal fluid, sweat, urine, milk, ascites, mucus, synovial fluid, peritoneal fluid, amniotic fluid, tissue, cells, and the like. The samples may be used directly as obtained from their respective sources or may be subjected to pretreatment and / or sample preparation workflows. For example, the samples may be pretreated by addition of an internal standard and / or dilution with another solution and / or mixing with a reagent, and the like. For example, the analytes of interest may generally be vitamin D, drugs of abuse, therapeutic agents, hormones, and metabolites. The internal standard sample may be a sample containing at least one internal standard at a known concentration. For further details regarding samples, see, for example, EP 3 425 369, the entire disclosure of which is incorporated herein by reference. Other analytes of interest are also possible.

[0025] The method includes the following steps, which may be performed in the given order by way of example. It should be noted, however, that different orders are also possible. Furthermore, one or more of the method steps may be performed only once or repeatedly. Furthermore, two or more of the method steps may be performed simultaneously or overlapped in time. The method may include further method steps not listed.

[0026] The method comprises the following steps: a) providing at least one mass spectrometry response curve by using at least one mass spectrometer; and b) evaluating the mass spectrometry response curve by using the at least one trained model to identify a start and an end of at least one peak in the mass spectrometry response curve. Including, The model is trained using a deep learning regression architecture.

[0027] As used herein, the term "mass spectrometry response curve" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning. The term can specifically refer to, but is not limited to, a one-dimensional representation of signal intensity. A mass spectrometry response curve has only one dimension. Specifically, the term "one-dimensional" can refer to the time axis and can refer to time being the only independent variable. Thus, as used herein, the term "one-dimensional" can refer to the fact that the only independent variable in the data is "time" and the dependent variable is "intensity". In particular, the present invention may not require two independent variables (e.g., "time" and "mass-to-charge ratio"), as is the case with some mass spectrometry data processing techniques. Typically, convolutional neural networks are used for two-dimensional data and applications, such as image recognition, where there are two independent variables (x- and y-directions of the image), as opposed to the use of only one-dimensional representations as proposed by the present invention. As used herein, the term "providing" is a broad term and should be given its ordinary and customary meaning to those skilled in the art, and should not be limited to a special or customized meaning. This term can specifically refer to, but is not limited to, the process of determining and / or generating and / or making available a mass spectrometry response curve, particularly by performing at least one measurement with a mass spectrometer. Thus, as used herein, the term "providing at least one mass spectrometry response curve by using at least one mass spectrometer" is a broad term and should be given its ordinary and customary meaning to those skilled in the art, and should not be limited to a special or customized meaning. This term can specifically refer to, but is not limited to, retrieving, particularly receiving, mass spectrometry response curve data obtained from a mass spectrometer, and / or performing at least one measurement with a mass spectrometer to determine mass spectrometry response curve data that can be used for further evaluation in step b).

[0028] As used herein, the term "evaluating a mass spectrometry response curve" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, the analysis of a mass spectrometry response curve. The evaluation may include identifying at least one peak and / or determining the start and end points of the peak and / or determining the peak area of ​​the peak. The evaluation may include applying at least one filter and / or using a background reduction technique and / or using at least one fitting route, etc. The evaluation may be performed using at least one evaluation device, as described in more detail below.

[0029] As used herein, the term "peak" of a mass spectrometry response curve is a broad term and should be given its ordinary and customary meaning to one of ordinary skill in the art and should not be limited to a special or customized meaning. The term can specifically refer to, but is not limited to, at least one maximum in a mass spectrometry response curve.

[0030] As used herein, the term "start" of a peak is a broad term and should be given its ordinary and customary meaning to one skilled in the art and should not be limited to a special or customized meaning. This term can specifically, but is not limited to, refer to the lower peak boundary. The start point can be a point on the time axis that defines the lower peak boundary. After the start point, the mass spectrometry response curve rises to a local maximum. The start point can be the beginning of the peak integration. As used herein, the term "end" of a mass spectrometry response curve is a broad term and should be given its ordinary and customary meaning to one skilled in the art and should not be limited to a special or customized meaning. This term can specifically, but is not limited to, refer to the upper peak boundary. The end point can be a point on the time axis that defines the upper peak boundary. The mass spectrometry response curve falls before reaching noise and / or background levels at the end point. The end point can be the end of the peak integration. The start and end points can be points on the time axis that are identified as the limits of the peak. The start and end point values ​​of the training data set, which will be described in more detail below, may be determined by manual evaluation by a trained user. The trained model may provide the start and end points of the further data. The peak area may typically be defined as the integral of the response curve between the start and end points.

[0031] As used herein, the term "peak identification" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term can specifically refer to, but is not limited to, a qualitative determination of a peak, such as the presence or absence of a peak, and / or a quantitative determination of a peak, such as determining the peak area of ​​the peak. Determining the peak area can include peak integration by using at least one mathematical operation and / or mathematical algorithm to determine, among other things, the peak area enclosed by the peak of a mass spectrometry response curve. Specifically, integrating a peak can include identifying and / or measuring the curve characteristics of a mass spectrometry response curve. Peak identification includes determining a start point and / or an end point. Peak identification can further include one or more of peak detection, peak finding, peak fitting, peak evaluation, background determination, and baseline determination. Peak integration can allow for the determination of one or more of peak area, retention time, peak height, and peak width. Peak identification can be automatic peak identification, i.e., peak identification performed by at least one computer and / or computer network and / or machine. In particular, automatic peak identification may be performed without manual action or interaction with a user.

[0032] As used herein, the term "trained model" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning. The term may specifically, but not exclusively, refer to a model for identifying peaks in mass spectrometry response curves that is trained on at least one training data set, also referred to as training data. In particular, the trained model is trained on existing data that has been pre-classified by experts. This can provide automated peak identification that is reliable and less susceptible to variance and error. The trained model may include an architecture and a set of weights for various filters or nodes defined by the architecture. The architecture of the CNN can reflect the complex relationship between the shape of the response curve and the location of the beginning and end of the peak.

[0033] The method can include at least one training step, in which the trained model is trained with at least one training data set. The training step can be offline training, while the peak identification in step b) of the proposed method can be online peak identification. Specifically, the training step can be performed before performing steps a) and b). As used herein, the term "online" is a broad term and should be given its ordinary and customary meaning to those skilled in the art, and should not be limited to a special or customized meaning. This term can specifically refer to, but is not limited to, during the measurement process by a mass spectrometer.

[0034] The model is trained using a deep learning regression architecture. As used herein, the term "deep learning" is a broad term and should be given its ordinary and customary meaning to those skilled in the art, and should not be limited to a special or customized meaning. The term may specifically, but is not limited to, refer to at least one method of machine learning. Deep learning may be based on at least one artificial neural network. As used herein, the term "deep learning regression architecture" is a broad term and should be given its ordinary and customary meaning to those skilled in the art, and should not be limited to a special or customized meaning. The term may specifically, but is not limited to, refer to a deep learning architecture configured to solve a regression problem.

[0035] The deep learning regression architecture may include a convolutional neural network. The convolutional neural network may be a multi-layer convolutional neural network. The convolutional neural network may comprise multiple convolutional layers. The convolutional layer is a one-dimensional layer, i.e. the convolution is applied in a one-dimensional time domain. In general, the convolutional layer is a standard basic building block in a convolutional neural network and is known per se to those skilled in the art. Mathematically, the convolutional layer corresponds to the operation of convolving input data with a convolutional kernel (see, for example, en.wikipedia.org / wiki / Convolutional_neural_network). As used herein, the term "convolutional layer" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, the operation of convolving input data with a one-dimensional convolutional kernel. The convolutional layer may be followed by multiple fully connected layers. A convolutional neural network can comprise multiple pooling layers. The structure of a convolutional neural network is well known to those skilled in the art, for example from en.wikipedia.org / wiki / Convolutional_neural_network#Convolutional. The use of convolutional neural networks for object recognition in images is well known to those skilled in the art. However, the present invention proposes a novel approach to use convolutional neural networks for 1D signal analysis.

[0036] Convolutional neural networks (CNNs) can be configured to solve regression problems. To solve regression problems, convolutional neural networks can include a regression layer as the final layer, in particular as opposed to a regular classification softmax layer. The regression layer can be a fully connected layer. The regression layer can have linear or sigmoid activation. Thus, the present invention proposes the use of one-dimensional convolutional neural networks in a regression framework. In particular, the present invention proposes the use of convolutional neural networks to fit complex functions that map inputs to peak locations. However, convolutional neural networks do not have to be used to classify 1D signals into categories.

[0037] The training step consists of the following sub-steps: i) providing at least one training data set comprising a plurality of input mass spectrometry response curves and corresponding ground truths; and ii) determining at least one model by using a deep learning regression architecture on the training data set; may include The sub-step of determining the model includes determining a model architecture and at least one parameter of the model.

[0038] Step i) may include providing more than 100, preferably more than 1000 input mass spectrometry response curves. For example, the model may be trained using 1270 vitamin D3 curves. The multiple input mass spectrometry response curves provided in step i) may be determined by performing multiple measurements using a mass spectrometer. For example, the multiple input mass spectrometry response curves provided in step i) may be or may include LC-MS data from a particular analyte, such as vitamin D2 or vitamin D3. As used herein, the term "correct answer" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. This term may specifically refer to, but is not limited to, the actual or true values ​​of the start and end points of the peaks of the corresponding input mass spectrometry response curves. The correct answer may indicate the location of the peak. For example, the correct answer may be the peak location or the start and end points of the peak. The correct answer may be provided by a trained LC-MS operator.

[0039] Injection into the mass spectrometer can result in four mass spectrometry response curves, two analyte mass spectrometry response curves and two internal standard mass spectrometry response curves. The training data set can be provided as a five-channel vector including an integration time vector, two analyte mass spectrometry response curves, and two internal standard mass spectrometry response curves. The raw time vector of the four response curves includes time steps that are at least slightly offset for each curve. The term "integration time vector" can refer to a time vector in which all four response curves are interpolated onto the same time grid. The training data set can be provided as input, also referred to as input data, to a convolutional neural network. For example, given an input mass spectrometry response curve of length N, the input can be a 2xN matrix, where the first row represents N intensity values, such as y values ​​of a chromatogram, and the second row represents N time values, such as x values ​​of a chromatogram. Integration can enable the convolutional neural network to propagate information between different mass spectrometry response curves, such that, for example, if a peak is very weak in a given analyte curve, the peak positions in other curves can be used to inform the position of the weak curve.

[0040] Training the model may include at least one normalization step. The normalization step may include normalizing the input data with respect to time. The normalization step may include transforming the time values ​​so that the expected retention time is located at t=0. The normalization step may include clipping the input data to a fixed time window around the expected retention time. The normalization step may include normalizing the intensity values ​​Y themselves by:

number

[0041] Training the model can include at least one extension step. To achieve generalization between analytes, the convolutional neural network can be trained with extended data that accounts for shift and scaling differences in the mass spectrometry response curve data. The extension step can allow for avoidance of overtraining. The extension step can include position extension and / or scale extension. Position extension can include shifting the peak position by a predetermined constant value. Position extension can include using a sliding window. This can account for the possibility of a peak becoming broadened or narrowed. Scaling extension can include scaling the peak by a predetermined value, such as 1.2. The training data set can be supplemented by randomly generating three new data sets using position extension and three new data sets using scale extension for each input mass spectrometry response curve.

[0042] The method may include using a deep learning regression architecture on the normalized and / or augmented training dataset and ground truth.

[0043] The deep learning regression architecture may be a convolutional neural network (CNN) built with Python's Keras library with TensorFlow as the backend. For information about Python's Keras library with TensorFlow, see https: / / www.tensorflow.org / , https: / / de.wikipedia.org / wiki / TensorFlow, and https: / / keras.io / or https: / / de.wikipedia.org / wiki / Keras. For example, the following settings may be used: As an optimizer, adaptive moment estimation (Adam) may be used. The loss function may be the mean squared error. The number of epochs may be 500, the batch size may be 16, and early stopping with patience may be 100.

[0044] The training step may further include at least one testing step using at least one test data set. The testing step may include validation of the trained model. As used herein, the term "test data set" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term may specifically refer to, but is not limited to, at least one mass spectrometry response curve and a corresponding ground truth that was not included in the training data set. The testing step may include using the determined model to determine the start and end points of at least one peak of the mass spectrometry response curve of the test data set. The performance of the determined model may be judged based on the determined start and end points and the ground truth of the mass spectrometry response curve of the test data set.

[0045] The method may comprise determining, in particular automatically, the peak area of ​​the peak of the mass spectrometric response curve provided in step i) by using the identified start and end points.

[0046] In a further aspect of the present invention, a computer program for identifying at least one peak in a mass spectrometry response curve is disclosed, which computer program comprises instructions that, when executed, cause a computer or a computer network to fully or partially execute steps a) and b) of a method according to the invention, such as a method according to any one of the embodiments disclosed above and / or any one of the embodiments disclosed in more detail below. Step a) may comprise a step that may be at least partially executed by a user, such as a step of sample preparation. However, embodiments are possible in which all steps of the method according to the invention are executed fully automatically. Thus, in particular, one, more than one or all of the method steps a) and b) as described above may be executed using a computer or a computer network, preferably using a computer program. In particular, the computer program may be stored on a computer-readable data carrier and / or a computer-readable storage medium.

[0047] Similarly, a computer-readable storage medium is disclosed, comprising instructions which, when executed, cause a computer or a computer network to fully or partially perform steps a) and b) of a method according to the present invention, such as a method according to any one of the embodiments disclosed above and / or any one of the embodiments disclosed in more detail below.

[0048] As used herein, the term "computer-readable storage medium" may specifically refer to a non-transitory data storage means, such as a hardware storage medium that stores computer-executable instructions. A computer-readable data carrier or storage medium may specifically be or include a storage medium, such as a random access memory (RAM) and / or a read-only memory (ROM).

[0049] Furthermore, the computer program may be embodied as a computer program product. As used herein, a computer program product may refer to a program as a tradeable product. The product may generally be present in any format, such as a paper format, or may be present on a computer-readable data carrier and / or a computer-readable storage medium. In particular, the computer program product may be distributed over a data network.

[0050] Further disclosed and proposed herein is a data carrier storing a data structure which, after being loaded into a computer or computer network, such as a working memory or main memory of a computer or computer network, is capable of performing the methods according to one or more of the embodiments disclosed herein.

[0051] Further disclosed and proposed herein is a computer program product having program code means stored on a machine-readable carrier for executing the method according to one or more of the embodiments disclosed herein when the program is executed on a computer or computer network. As used herein, a computer program product refers to a program as a tradeable product. The product can generally be present in any format, such as a paper format, or can be present on a computer-readable data carrier and / or a computer-readable storage medium. In particular, the computer program product may be distributed over a data network.

[0052] Finally, further disclosed and suggested herein is a modulated data signal that includes instructions readable for a computer system or computer network to carry out a method according to one or more of the embodiments disclosed herein.

[0053] With respect to computer-implemented aspects of the present invention, one or more or all of the method steps of the method according to one or more of the embodiments disclosed herein can be performed by using a computer or a computer network.Thus, in general, any of the method steps that include providing and / or manipulating data can be performed by using a computer or a computer network.In general, these method steps can include any method steps, except for those that typically require manual work, such as providing a sample and / or performing the actual measurement in certain aspects.

[0054] Specifically, the following is further disclosed herein: - a computer or computer network comprising at least one processor, the processor being configured to execute a method according to one of the embodiments described in this specification. - a data structure loadable into a computer and configured, when executed on the computer, to perform a method according to one of the embodiments described in this specification. - a computer program configured to carry out the method according to one of the embodiments described herein when it is run on a computer. - a computer program comprising program means for carrying out the method according to one of the embodiments described herein when said computer program is run on a computer or on a computer network. A computer program comprising program means according to the preceding embodiment, the program means being stored on a computer readable storage medium. - A storage medium having a data structure stored thereon, the storage medium being configured to perform a method according to one of the embodiments described in this specification after the data structure has been loaded into a main storage and / or working storage of a computer or computer network. - a computer program product having program code means for performing a method according to one of the embodiments described herein when the program code means is executed on a computer or on a computer network, the program code means being storable or stored on a storage medium.

[0055] In a further aspect of the present invention, an apparatus for monitoring at least one analyte in a sample is disclosed. The apparatus comprises: - at least one mass spectrometer configured to provide at least one mass spectrometric response curve; - at least one evaluator configured to evaluate the mass spectrometry response curve by using the at least one trained model, thereby identifying a start point and an end point of at least one peak in the mass spectrometry response curve; Equipped with The model is trained using a deep learning regression architecture.

[0056] The apparatus may be configured to perform a method for identifying at least one peak in a mass spectrometry response curve according to any one of the preceding embodiments relating to the method. Regarding the definitions of the features of the apparatus and any features of the apparatus, reference may be made to one or more of the embodiments of the method disclosed above or disclosed in more detail below.

[0057] As generally used herein, the term "evaluation device" is a broad term and should be given its ordinary and customary meaning to those skilled in the art and should not be limited to a special or customized meaning. The term can specifically refer to any device configured to perform the recited operations, but is not limited thereto. The evaluation device can include at least one processing unit. The processing unit can be any logic circuit configured to perform the basic operations of a computer or system, and / or generally a device configured to perform calculations or logical operations. In particular, the processing unit can be configured to process basic instructions that drive the computer or system. By way of example, the processing unit can include at least one arithmetic logic unit (ALU), at least one floating point unit (FPU), such as a numeric coprocessor or numeric coprocessor, a number of registers, specifically registers configured to provide operands to the ALU and store the results of the operations, and memories, such as L1 and L2 cache memories. In particular, the processing unit can be a multi-core processor. In particular, the processing unit can be or include a central processing unit (CPU). Additionally or alternatively, the processing unit may be or comprise a microprocessor, and thus, in particular, the elements of the processing unit may be included in one single integrated circuit (IC) chip. Additionally or alternatively, the processing unit may be or comprise one or more application specific integrated circuits (ASICs) and / or one or more field programmable gate arrays (FPGAs), etc. The processing unit may be, in particular, configured, such as by software programming, to perform one or more evaluation operations.

[0058] The assessment device may be configured to carry out step b) of the method according to the invention as described in detail above or in more detail below. The assessment device may further be configured to carry out a training step as described in detail above or in more detail below.

[0059] In summary, without excluding further embodiments, the following embodiments can be envisaged:

[0060] Embodiment 1: A computer-implemented method for identifying at least one peak in a mass spectrometry response curve, comprising the steps of: a) providing at least one mass spectrometry response curve by using at least one mass spectrometer; and b) evaluating the mass spectrometry response curve by using the at least one trained model to identify a start and an end of at least one peak in the mass spectrometry response curve. Contains The model is trained using a deep learning regression architecture.

[0061] Embodiment 2: The method according to the preceding embodiment, wherein the deep learning regression architecture comprises a convolutional neural network.

[0062] Embodiment 3: The method according to the preceding embodiment, wherein the convolutional neural network is a multi-layer convolutional neural network.

[0063] Embodiment 4: The method according to the preceding embodiment, wherein the convolutional neural network comprises a plurality of convolutional layers, the convolutional layers being one-dimensional layers.

[0064] Embodiment 5: A method according to any one of the two preceding embodiments, wherein the convolutional neural network comprises a recurrent layer as a final layer, the recurrent layer having linear or sigmoid activation.

[0065] Embodiment 6: The method according to any one of the preceding three embodiments, wherein the convolutional neural network comprises a plurality of pooling layers.

[0066] Embodiment 7: A method comprising at least one training step, The training step consists of the following sub-steps: i) providing at least one training data set comprising a plurality of input mass spectrometry response curves and corresponding ground truths; and ii) determining at least one model by using a deep learning regression architecture on the training data set; Including, 13. A method according to any one of the preceding embodiments, wherein the substep of determining the model includes determining a model architecture and at least one parameter of the model.

[0067] Embodiment 8: A method according to the preceding embodiment, wherein the training data set is provided as a five-channel vector including an integration time vector, two analyte mass analytical response curves, and two internal standard mass analytical response curves.

[0068] Embodiment 9: A method according to any one of the previous two embodiments, wherein the training of the model includes at least one normalization step and / or at least one augmentation step.

[0069] Embodiment 10: A method according to any one of the preceding two embodiments, wherein the training step further includes at least one testing step using at least one test data set, the testing step including determining a start point and an end point of at least one peak of a mass spectrometric response curve of the test data set using the determined model, and the performance of the determined model is judged based on the determined start point and end point and a ground truth of the mass spectrometric response curve of the test data set.

[0070] Embodiment 11: Determining the peak area of ​​a peak in a mass spectrometry response curve by using the identified start and end points 3. The method according to any one of the preceding embodiments, comprising:

[0071] Embodiment 12: A computer program for identifying at least one peak in a mass spectrometry response curve, configured to cause a computer or a computer network to fully or partially perform a method for identifying at least one peak in a mass spectrometry response curve according to any one of the preceding embodiments relating to the method, when executed on the computer or on a computer network, comprising: A computer program configured to perform at least step b) of the method for identifying at least one peak in a mass spectrometry response curve according to any one of the preceding embodiments relating to the method.

[0072] Embodiment 13: A computer program product having program code means, comprising: A computer program product, wherein the program code means is storable or stored on a storage medium for performing at least step b) of a method for identifying at least one peak in a mass spectrometry response curve according to any one of the preceding embodiments relating to the method, when said program code means is executed on a computer or a computer network.

[0073] Embodiment 14: A device for monitoring at least one analyte in a sample, comprising: - at least one mass spectrometer configured to provide at least one mass spectrometric response curve; - at least one evaluator configured to evaluate the mass spectrometry response curve by using the at least one trained model, thereby identifying a start point and an end point of at least one peak in the mass spectrometry response curve; Equipped with The model is trained using a deep learning regression architecture.

[0074] Embodiment 15: An apparatus according to the preceding embodiments, configured to perform a method for identifying at least one peak in a mass spectrometry response curve as described in any one of the preceding embodiments relating to the method.

[0075] Further optional features and embodiments are disclosed in more detail in the following description of the embodiments, preferably in conjunction with the dependent claims. Therein, each optional feature may be realized in an independent manner as well as in any possible combination, as understood by a person skilled in the art. The scope of the present invention is not limited by the preferred embodiments. The embodiments are illustrated diagrammatically in the drawings, in which the same reference numbers in these figures refer to the same or functionally equivalent elements.

[0076] In the drawings: [Brief description of the drawings]

[0077] [Figure 1] 1 illustrates an embodiment of a computer-implemented method for identifying at least one peak in a mass spectrometry response curve in accordance with the present invention. [Diagram 2] 1 illustrates an embodiment of a device for monitoring at least one analyte in a sample according to the present invention. [Diagram 3] 1 illustrates an embodiment of a deep learning regression architecture. [Figure 4] 4 illustrates an embodiment of a training step according to the present invention. [Figure 5A] 1 illustrates an embodiment of normalization and cropping. [Figure 5B] 1 illustrates an embodiment of normalization and cropping. [Figure 5C] 1 illustrates an embodiment of normalization and cropping. [Figure 6A] 1 illustrates an embodiment of expansion and scaling. [Figure 6B] 1 illustrates an embodiment of expansion and scaling. [Figure 6C] 1 illustrates an embodiment of expansion and scaling. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0078] 1 shows, in a highly schematic manner, a flow chart of one embodiment of a computer-implemented method 110 for identifying at least one peak in a mass spectrometry response curve according to the present invention. The method comprises the following steps: a) providing at least one mass spectrometry response curve by using at least one mass spectrometer 114 (depicted by reference numeral 112); and b) identifying a start and an end of at least one peak in the mass spectrometry response curve by evaluating the mass spectrometry response curve by using the at least one trained model (depicted with reference numeral 116); may include The model is trained using a deep learning regression architecture 118.

[0079] FIG. 2 shows an embodiment of an apparatus 120 for monitoring at least one analyte in a sample according to the present invention, comprising a mass analyzer 114. The mass analyzer 114 may be configured to detect at least one analyte based on mass-to-charge ratio. The mass analyzer 114 may be or may comprise at least one quadrupole analyzer comprising at least one quadrupole 122 as a mass filter. The quadrupole mass analyzer may comprise multiple quadrupoles. For example, the quadrupole mass analyzer may be a triple quadrupole mass analyzer. The quadrupole mass analyzer may comprise at least one power supply circuit (not shown here) configured to apply at least one direct current (DC) voltage and at least one alternating current (AC) voltage between two pairs of electrodes of the quadrupole. The power supply circuit may be configured to hold each opposing electrode pair at the same potential. The power supply circuit may be configured to periodically change the sign of the charge of the electrode pairs to allow stable trajectories only for ions within a certain mass-to-charge ratio m / z range. The trajectories of the ions in the mass filter can be described by the Mathieu differential equation: To measure ions of different m / z values, the DC and AC voltages can be varied over time to allow ions with different m / z values ​​to be transmitted to the detector 124 of the mass analyzer 114.

[0080] The mass analyzer 114 may further include at least one ionization source 126. The ionization source 126 may be configured to generate ions from, for example, neutral gas molecules. The ionization source 126 may be or may include at least one source selected from the group consisting of at least one gas phase ionization source, such as at least one electron impact (EI) source or at least one chemical ionization (CI) source, at least one desorption ionization source, such as at least one plasma desorption (PDMS) source, at least one fast atom bombardment (FAB) source, at least one secondary ion mass spectrometry (SIMS) source, at least one laser desorption (LDMS) source, and at least one matrix-assisted laser desorption (MALDI) source, at least one spray ionization source, such as at least one thermospray (TSP) source, at least one atmospheric pressure chemical ionization (APCI) source, at least one electrospray (ESI) source, and at least one atmospheric pressure ionization (API) source.

[0081] The mass analyzer 114 may include a detector 124. The detector may be configured to detect incoming ions. The detector 124 may be configured to detect charged particles. The detector 124 may be or may include at least one electron multiplier.

[0082] The mass analyzer 114, in particular the detector 124 and / or at least one evaluation device 128 of the mass analyzer 114, can be configured to determine at least one mass spectrum of the detected ions. The mass spectrum can be an image made up of pixels. The at least one evaluation device 128 can identify analytes in the sample that can integrate the signal detected by the detector in a certain m / z range to determine the intensity of the pixels of the resulting mass spectrum. In particular, the evaluation device 128 can be configured to correlate a known mass to the identified mass or can be configured by a characteristic fragmentation pattern.

[0083] The mass analyzer 114 may be or comprise a liquid chromatography mass analyzer. The mass analyzer 114 may be connected to and / or comprise at least one liquid chromatograph. A liquid chromatograph may be used for sample preparation for the mass analyzer 114. Other embodiments of sample preparation may be possible, such as at least one gas chromatograph. The mass analyzer 114 may comprise at least one liquid chromatograph. The liquid chromatography mass analyzer may be or comprise at least one high performance liquid chromatography (HPLC) device or at least one micro liquid chromatography (μLC) device. The liquid chromatography mass analyzer may comprise a liquid chromatography (LC) device and a mass spectrometry (MS) device, which in this case is a mass filter, the LC device and the mass filter being coupled via at least one interface. The interface coupling the LC device and the MS device can include an ionization source configured to generate molecular ions and transport the molecular ions to the gas phase. The interface can further include at least one ion mobility module disposed between the ionization source and the mass filter. For example, the ion mobility module can be a high field asymmetric waveform ion mobility spectrometry (FAIMS) module.

[0084] The LC device can be configured to separate one or more analytes of interest of the sample from other components of the sample for detection of the one or more analytes by the mass spectrometry device 114. The LC device can include at least one LC column. For example, the LC device can be a single-column LC device or a multi-column LC device having multiple LC columns. The LC column can have a stationary phase through which a mobile phase is pumped for separation and / or elution and / or transport of the analytes of interest. The liquid chromatography mass spectrometry device can further include a sample preparation station for automated pre-treatment and preparation of samples, each of which includes at least one analyte of interest.

[0085] The sample may be any test sample, such as a biological sample and / or an internal standard sample. The sample may contain one or more analytes of interest. For example, the test sample may be selected from the group consisting of physiological fluids including blood, serum, plasma, saliva, ocular lens fluid, cerebrospinal fluid, sweat, urine, milk, peritoneal fluid, mucus, synovial fluid, peritoneal fluid, amniotic fluid, tissue, cells, and the like. The samples may be used directly as obtained from their respective sources or may be subjected to pretreatment and / or sample preparation workflows. For example, the samples may be pretreated by addition of an internal standard and / or dilution with another solution and / or mixing with a reagent, and the like. For example, the analytes of interest may generally be vitamin D, drugs of abuse, therapeutic agents, hormones, and metabolites. The internal standard sample may be a sample containing at least one internal standard at a known concentration. For further details regarding samples, see, for example, EP 3 425 369 A1, the entire disclosure of which is incorporated herein by reference. Other analytes of interest are also possible.

[0086] The mass spectrometry response curve provided in step a) 112 may be a one-dimensional representation of the signal intensity. The mass spectrometry response curve only has one dimension. The mass spectrometry response curve may be provided by determining and / or generating and / or making available the mass spectrometry response curve, in particular by performing at least one measurement with the mass analyzer 114.

[0087] The evaluation of the mass spectrometry response curve in step b) 116 can include performing at least one analysis of the mass spectrometry response curve. The evaluation 116 can include identifying at least one peak and / or determining the start and end points of the peak and / or determining the peak area of ​​the peak. The evaluation 116 can include applying at least one filter and / or using a background reduction technique and / or using at least one fitting route, etc.

[0088] The peak of the mass spectrometry response curve may be at least one local maximum of the mass spectrometry response curve. The start of the peak may be a lower peak boundary. The start may be a point on the time axis that defines the lower peak boundary. After the start, the mass spectrometry response curve rises to a local maximum. The start may be a start of the peak integration. The end of the mass spectrometry response curve may be an upper peak boundary. The end may be a point on the time axis that defines the upper peak boundary. The mass spectrometry response curve falls before reaching the noise and / or background level at the end. The end may be an end of the peak integration. The start and end may be points on the time axis that are identified as the limits of the peak. The start and end values ​​of the training data set, which are described in more detail below, may be determined by manual evaluation by a trained user. The trained model may provide the start and end of further data. The peak area may typically be defined as the integral of the response curve between the start and end points.

[0089] Step b) 116 may include peak identification. Peak identification may include a qualitative determination of a peak, such as the presence or absence of a peak, and / or a quantitative determination of a peak, such as the determination of the peak area of ​​the peak. The determination of the peak area may include peak integration by using at least one mathematical operation and / or mathematical algorithm to determine, among other things, a peak area enclosed by a peak of the mass spectrometry response curve. In particular, peak integration may include identifying and / or measuring curve characteristics of the mass spectrometry response curve. Peak identification includes the determination of a start point and / or an end point. Peak identification may further include one or more of peak detection, peak finding, peak fitting, peak evaluation, background determination, and baseline determination. Peak integration may allow the determination of one or more of peak area, retention time, peak height, and peak width. Peak identification may be automatic peak identification, i.e. peak identification performed by at least one computer and / or computer network and / or machine. In particular, automatic peak identification may be performed without manual action or user interaction.

[0090] The trained model used in step b) 116 may be or may comprise a model for identifying peaks in mass spectrometry response curves trained on at least one training data set, also called training data. In particular, the trained model is trained on existing data previously classified by experts. This allows to provide an automated peak identification that is reliable and less susceptible to variance and errors. The trained model may include an architecture and a set of weights for different filters or nodes defined by the architecture. The architecture of the CNN may reflect the complex relationship between the shape of the response curve and the location of the start and end of the peak.

[0091] The method 110 may include at least one training step 130, in which the trained model is trained on at least one training dataset. The training step 130 may be offline training, whereas the peak identification in step b) 116 of the proposed method may be online peak identification. In particular, the training step 130 may be performed before performing steps a) 112 and b) 116.

[0092] The model is trained using a deep learning regression architecture 118. The deep learning regression architecture 118 may be a deep learning architecture configured to solve a regression problem.

[0093] The deep learning regression architecture 118 may include a convolutional neural network. The convolutional neural network may be a multi-layer convolutional neural network. FIG. 3 illustrates an embodiment of the deep learning regression architecture 118. The convolutional neural network may include at least one feature trainer 132. The feature trainer may include multiple convolutional layers 134. The convolutional layers 134 are one-dimensional layers, i.e. the convolution is applied in one-dimensional time domain. Each of the convolutional layers 134 may include at least one rectified linear unit, also referred to as ReLU. The convolutional neural network 118 may include multiple pooling layers 136. The structure of a convolutional neural network is well known to those skilled in the art, for example, from https: / / en.wikipedia.org / wiki / Convolutional_neural_network#Convolutional. The use of convolutional neural networks for object recognition in images is well known to those skilled in the art. However, the present invention proposes a novel approach that uses convolutional neural networks for 1D signal analysis.

[0094] The convolutional neural network 118 can be configured to solve a regression problem. To solve a regression problem, the convolutional neural network 118 can include a regression layer 138 as the final layer, in particular as opposed to a regular classification softmax layer. The convolutional neural network 118 can include at least one smoothing layer 140.

[0095] The regression layer 138 may be a fully connected layer 139. The regression layer 138 may have linear or sigmoid activation. Thus, the present invention proposes the use of a one-dimensional convolutional neural network in a regression framework. Specifically, the present invention proposes using the convolutional neural network 118 to fit a complex function that maps inputs to peak locations. However, the convolutional neural network 118 does not have to be used to classify 1D signals into categories.

[0096] 4 illustrates one embodiment of the training step 130. The training step includes the following sub-steps: i) providing at least one training data set (denoted by reference numeral 142) comprising a plurality of input mass spectrometry response curves and corresponding ground truths; and ii) determining at least one model by using the deep learning regression architecture 118 on the training data set (depicted with reference numeral 144); Including, The sub-step of determining the model includes determining a model architecture and at least one parameter of the model.

[0097] Step i) 142 may include providing more than 100, preferably more than 1000, input mass spectrometry response curves. For example, the model may be trained using 1270 vitamin D3 curves. The multiple input mass spectrometry response curves provided in step i) 142 may be determined by performing multiple measurements using a mass spectrometer. For example, the multiple input mass spectrometry response curves provided in step i) 142 may be or may include LC-MS data from a particular analyte, such as vitamin D2 or vitamin D3. The correct answer may be the actual or true value of the start and end of the peak of the corresponding input mass spectrometry response curve. The correct answer may indicate the location of the peak. The correct answer may be provided by a trained LC-MS operator.

[0098] Injection into the mass analyzer 114 can result in four mass spectrometry response curves, namely two analyte mass spectrometry response curves and two internal standard mass spectrometry response curves. The training data set can be provided as a five-channel vector including an integration time vector, two analyte mass spectrometry response curves, and two internal standard mass spectrometry response curves. The raw time vector of the four response curves includes time steps that are at least slightly offset for each curve. The integration time vector can be a time vector in which all four response curves are interpolated onto the same time grid. The training data set can be provided as input, also referred to as input data, to a convolutional neural network. For example, given an input mass spectrometry response curve of length N, the input can be a 2xN matrix, where the first row represents N intensity values, such as y values ​​of a chromatogram, and the second row represents N time values, such as x values ​​of a chromatogram. Integration can enable the convolutional neural network to propagate information between different mass spectrometry response curves, such that, for example, if a peak is very weak in a given analyte curve, the peak positions in other curves can be used to inform the position of the weak curve.

[0099] Training the model 130 may include at least one normalization step 146. The normalization step 146 may include normalizing the input data with respect to time. The normalization step 146 may include transforming the time values ​​so that the predicted retention time is located at t=0 in FIG. 5B. The normalization step 146 may include clipping the input data to a fixed time window around the predicted retention time. The normalization step 146 may include normalizing the intensity values ​​Y themselves by:

number

[0100] 5A-5C show an embodiment of normalization and clipping. Fig. 5A shows intensity I as a function of time (in minutes) for the original input data. Fig. 5B shows the input data after normalization with respect to time, and Fig. 5C shows the input data after further normalization with respect to intensity and clipping. Circle 148 indicates the correct answer for the start of the peak, and circle 150 represents the correct answer for the location of the end of the peak.

[0101] Training the model 130 may include at least one augmentation step 152. To achieve generalization between analytes, the convolutional neural network may be trained with augmented data that accounts for shift and scaling differences in the mass spectrometry response curve data. The augmentation step 152 may allow for overtraining to be avoided. The augmentation step 152 may include position augmentation and / or scale augmentation. Position augmentation may include shifting the peak position by a predetermined constant value. Position augmentation may include using a sliding window. This may allow for the possibility of a peak becoming broader or narrower. Scaling augmentation may include scaling the peak by a predetermined value, such as 1.2. The training data set may be supplemented by randomly generating three new data sets using position augmentation and three new data sets using scale augmentation for each input mass spectrometry response curve. FIG. 6B shows the original input data with ground truths 148 and 150. FIG. 6A shows position augmentation, depicting the original curve 154 and the augmented curve 156. FIG. 6C shows scale augmentation by 1.2.

[0102] The deep learning regression architecture 118 may be a convolutional neural network (CNN) built with Python's Keras library with TensorFlow as a backend. For information about the Keras library for Python with TensorFlow, see https: / / www.tensorflow.org / , https: / / de.wikipedia.org / wiki / TensorFlow, and https: / / keras.io / or https: / / de.wikipedia.org / wiki / Keras. For example, the following settings may be used: As an optimizer, adaptive moment estimation (Adam) may be used. The loss function may be the mean squared error. The number of epochs may be 500, the batch size may be 16, and early stopping with patience may be 100.

[0103] The training step 130 may include using a deep learning regression architecture 118 on the normalized and / or augmented training data set and the ground truth. As an output 158, the model architecture and at least one parameter of the model may be provided.

[0104] The training step 130 may further include at least one testing step 160 using at least one test data set. The testing step 160 may include validation of the trained model. The test data set may include at least one, preferably a plurality of mass spectrometry response curves and corresponding ground truths that were not included in the training data set. The testing step 160 may include determining, using the determined model, the start and end points of at least one peak of the mass spectrometry response curve of the test data set. The performance of the determined model may be judged based on the determined start and end points and the ground truths of the mass spectrometry response curve of the test data set.

[0105] In an experimental setup, the deep learning regression architecture 118 was trained on 1462 curves, especially fragment curves, of chromatograms from 488 unique samples. The chromatograms contained the correct answer. For the testing step 160, 10% or 49 samples were left for final validation. For the testing step 160, 439 samples from the 488 samples were used to train the algorithm for cross-validation. For cross-validation, the 439 samples were divided into five groups. Four of the five were used for training 130 and one for testing 160. This is done for all possible permutations. It was found that the method 110 according to the invention generalizes well between analytes and improves the accuracy of LC-MS promising to improve the reliability of automatic peak identification.

[0106] The following table shows the performance (R of peak locations) when the model is trained on a particular analyte and measurement system (denoted as Extended Data System 2) and then applied to another analyte and measurement system (denoted as Extended Data System 1). 2 ) The table includes similar analytes, namely Vitamin D2 and Vitamin D3, as well as a completely different substance, testosterone.

[0107] [Table 1] [Explanation of symbols]

[0108] 110 Method 112 Step a) 114 Mass spectrometer 116 Step b) 118 Deep Learning Regression Architectures 120 equipment 122 Quadrupole 124 Detector 126 Ionization Source 128 Evaluation Equipment 130 Training Steps 132 Feature Learning Unit 134 Convolutional Layer 136 Pooling Layer 138 Recurrent Layer 139 Fully connected layer 140 Planarization layer 142 Step i) 144 Step ii) 146 Normalization Steps 148 yen 150 yen 152 Extended Steps 154 Original curve 156 Extended Curves 158 Output 160 Inspection Steps

Claims

1. 1. A computer-implemented method (110) for identifying at least one peak in a mass spectrometry response curve, comprising: a) providing (112) at least one mass spectrometry response curve by using at least one mass spectrometer (114); b) evaluating the mass spectrometry response curve having a one-dimensional representation of signal intensity by using at least one trained model to identify a start and an end of at least one peak in the mass spectrometry response curve (116); Contains The method (110), wherein the model is trained using a deep learning regression architecture (118).

2. 2. The method of claim 1 , wherein the deep learning regression architecture comprises a convolutional neural network.

3. 3. The method of claim 2, wherein the convolutional neural network is a multi-layer convolutional neural network.

4. 4. The method (110) of claim 3, wherein the convolutional neural network comprises a plurality of convolutional layers (134), the convolutional layers (134) being one-dimensional layers.

5. 5. The method (110) of claim 3 or 4, wherein the convolutional neural network comprises a recurrent layer (138) as a final layer, the recurrent layer (138) having a linear or sigmoid activation.

6. At least one training step (130), The training step includes: i) providing at least one training data set comprising a plurality of input mass spectrometry response curves and corresponding ground truths (142); ii) determining (144) at least one model by using said deep learning regression architecture (118) on said training data set; Including, The method (110) according to any one of the preceding claims, wherein the sub-step of determining the model comprises determining a model architecture and at least one parameter of the model.

7. 7. The method (110) of claim 6, wherein the training data set is provided as a five-channel vector including an integration time vector, two analyte mass spectrometric response curves, and two internal standard mass spectrometric response curves.

8. The method (110) of claim 6 or 7, wherein the training (130) of the model comprises at least one normalization step (146) and / or at least one expansion step (152).

9. 9. The method of claim 7 or 8, wherein the training step (130) further comprises at least one testing step (160) using at least one test data set, the testing step (160) comprising determining a start and end point of at least one peak of the mass spectrometry response curve of the test data set using the determined model, and the performance of the determined model is judged based on the determined start and end points and a ground truth of the mass spectrometry response curve of the test data set.

10. determining a peak area of ​​said peak in said mass spectrometry response curve by using said identified start and end points; The method (110) of any one of claims 1 to 9, comprising:

11. A computer program for identifying at least one peak in a mass spectrometry response curve, configured to cause a computer or a computer network to fully or partially execute a method (110) for identifying at least one peak in a mass spectrometry response curve according to any one of claims 1 to 10 relating to the method, said computer program comprising: A computer program configured to perform at least step b) of the method (110) for identifying at least one peak in a mass spectrometry response curve according to any one of claims 1 to 10.

12. 1. A computer program product having program code means, comprising:

11. A computer program product, said program code means being storable or stored on a storage medium for performing at least step b) of the method (110) for identifying at least one peak in a mass spectrometry response curve according to any one of claims 1 to 10, when said program code means is executed on a computer or a computer network.

13. 1. An apparatus (120) for monitoring at least one analyte in a sample, comprising: at least one mass analyzer (114) configured to provide at least one mass spectrometry response curve; at least one evaluation device (128) configured to evaluate said mass spectrometry response curve having a one-dimensional representation of signal intensity by using at least one trained model, thereby identifying the start and end of at least one peak of said mass spectrometry response curve; Equipped with The model is trained using a deep learning regression architecture (118), the apparatus (120).

14. An apparatus (120) according to claim 13, configured to carry out a method for identifying at least one peak in a mass spectrometry response curve according to any one of the methods claims 1 to 10.

Images