JP2021174596A - Sweet spot prediction method and sweet spot prediction device - Google Patents

Abstract

To easily predict the position of a sweet spot in MALDI mass spectrometry.SOLUTION: Learning data in which an optical image obtained by photographing an adhesion region of a predetermined sample on a sample plate and information indicating the position of a sweet spot in the adhesion region of the predetermined sample are associated with each other is acquired, a prediction model that predicts the position of the sweet spot is generated by machine learning using the learning data, and a prediction model is applied to the optical image obtained by photographing the adhesion region of an unknown sample to predict the position of the sweet spot within the adhesion region of the unknown sample.SELECTED DRAWING: Figure 6

Description

The present invention relates to a sweet spot prediction method and a sweet spot prediction device.

The MALDI (Matrix Assisted Laser Desorption / Ionization) method is well known as one of the ionization methods for mass spectrometers. In the MALDI method, in order to analyze a sample that does not easily absorb laser light or a sample that is easily damaged by laser light such as protein, a substance that easily absorbs laser light and is easily ionized is mixed with the sample in advance as a matrix. This is a method of ionizing the compound contained in the sample by irradiating it with laser light. A mass spectrometer (that is, a MALDI mass spectrometer) equipped with an ion source for ionizing a sample by the MALDI method can analyze a high molecular weight polymer compound without causing much cleavage, and has detection sensitivity. In recent years, it has been widely used in fields such as life science because it is suitable for microanalysis due to its high molecular weight.

In mass spectrometry by the MALDI method (that is, MALDI mass spectrometry), for example, a target sample is formed in a shallow bowl-shaped recess called a well formed on a metal plate called a sample plate (or target plate). Samples are prepared by dropping a mixed solution with a matrix, drying it, and crystallizing it.

Normally, the shape of one well is a circular shape in the top view, but the region occupied by the sample prepared above (strictly speaking, a mixed crystal of the target sample and the matrix) in the well (hereinafter referred to as the sample region). The distribution of the target sample in (called) is not always uniform. Therefore, there are many target samples in the sample region, and the optimum site for measurement (hereinafter referred to as “sweet spot”) is often deviated from the center of the sample region.

In order to perform highly sensitive and highly accurate analysis in mass spectrometry by the MALDI method, it is preferable to irradiate the sweet spot of each sample with a laser beam and subject the ions generated thereby to mass spectrometry. However, for the above reasons, the position of the sweet spot varies widely from well to well, and it is difficult to determine where the sweet spot is by visual observation through a microscope.

Therefore, the conventional MALDI mass spectrometer is equipped with a function called raster measurement. In raster measurement, laser beam irradiation is sequentially performed a plurality of times (for example, tens to hundreds of times) for each of a large number of measurement points preset at different positions in a sample region in one well. This is a measurement method in which the final measurement data is derived by acquiring the measurement data and integrating all the measurement data.

However, usually, a large number of wells are provided on the sample plate, and it takes a considerable amount of time to perform raster measurement on all of the sample regions formed in each well. Therefore, it is difficult to increase the analysis throughput in raster measurement.

Therefore, before acquiring the actual measurement data, the measurement data is acquired by preliminarily irradiating each of the large number of measurement points set in each well with laser light several times. It is also performed to search for a sweet spot by comparing the signal intensities obtained by such preliminary measurement (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2012-23801 ([0004])

However, since it is necessary to irradiate a large number of measurement points with laser even in the preliminary measurement, it takes some time to find the sweet spot. In addition, at one measurement point, even if the ionic strength measured by several laser irradiations is low, the ionic strength may increase as the laser irradiation is repeated, but in the preliminary measurement, the number of laser irradiations is high. Due to the small number, such a measurement point could not be found as a sweet spot. In addition, since the target sample is consumed during the preliminary measurement, sufficient signal strength may not be obtained when the actual measurement is performed thereafter.

The present invention has been made in view of the above points, and an object of the present invention is to provide a sweet spot prediction device and a sweet spot prediction method capable of easily predicting the position of a sweet spot in MALDI mass spectrometry. To provide.

The method for predicting sweet spots according to the present invention, which has been made to solve the above problems, is
Learning data in which the optical image obtained by photographing the adhesion region of the predetermined sample on the sample plate and the information indicating the position of the sweet spot in the adhesion region of the predetermined sample are associated with each other is acquired.
A prediction model that predicts the position of the sweet spot is generated by machine learning using the training data.
By applying the prediction model to an optical image obtained by photographing an adhesion region of an unknown sample, the position of a sweet spot in the adhesion region of the unknown sample is predicted.

Further, the sweet spot prediction device according to the present invention made to solve the above problems is
A learning data acquisition unit that acquires learning data in which an optical image obtained by photographing an adhesion region of a predetermined sample on a sample plate and information indicating the position of a sweet spot in the adhesion region of the predetermined sample are associated with each other. When,
A prediction model generation unit that generates a prediction model that predicts the position of the sweet spot by executing machine learning using the training data.
A prediction unit that predicts the position of the sweet spot in the adhesion region of the unknown sample by applying the prediction model to the optical image obtained by photographing the adhesion region of the unknown sample.
It has.

According to the sweet spot prediction method or the sweet spot prediction device according to the present invention, the position of the sweet spot in MALDI mass spectrometry can be easily predicted.

The schematic block diagram of the MALDI mass spectrometry system which concerns on one Embodiment of this invention. The block diagram which shows the composition of the main part of the control / processing part in the same embodiment. The flowchart which shows the procedure of the process concerning the generation of the prediction model in the same embodiment. The figure which shows an example of the sample image for learning. The figure which shows an example of a heat map image. The flowchart which shows the procedure of the process concerning the prediction of the sweet spot position and the execution of mass spectrometry in the same embodiment. The flowchart which shows another example of the procedure of the process concerning the prediction of the sweet spot position and the execution of mass spectrometry.

Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram of a MALDI mass spectrometry system including a sweet spot prediction device according to the present embodiment.

This MALDI mass spectrometry system includes a mass spectrometer 10 and a control / processing device 50, and the control / processing device 50 corresponds to the "sweet spot prediction device" in the present invention.

The mass spectrometer 10 takes an image of an ion source 20 for generating ions, a mass spectrometer 30 that separates and detects the ions based on m / z (mass-to-charge ratio), and an optical image of a sample. A unit 40 and the like are provided.

The ion source 20 collects on the sample stage 22 on which the sample plate 21 is placed on the upper surface thereof, the laser emitting portion 26 that emits the laser light, and the sample S that reflects the laser light and is formed on the sample plate 21. The first reflecting mirror 27 that shines, the aperture plate 24 that is provided above the sample stage 22 and shields diffused ions, and the ions that have passed through the aperture provided on the aperture plate 24 are transported to the mass analysis unit 30. The ion transport optical system 25 and the like are included.

The sample stage 22 can be moved in the biaxial directions of the X-axis and the Y-axis in FIG. 1 (that is, the spreading direction of the plate surface of the sample plate 21) by the stage driving unit 23 including the motor and the like.

A large number of circular dents called wells are two-dimensionally formed on the upper surface of the sample plate 21, and a sample solution (a mixed solution of a target sample and a matrix) is formed in each of the wells. ) Is dropped and dried to form a solid sample S in each well.

The observation image of the sample S is introduced into the photographing unit 40 provided with the CCD camera or the like via the second reflecting mirror 28, and the sample image formed by the photographing unit 40 is sent to the control / processing device 50.

The mass spectrometer 30 is a reflector-type TOFMS including a free flight space 31 in which ions freely fly without the influence of an electric field and a reflector 32 in which ions are turned back by the action of a DC electric field, and the end of the flight. An ion detector 33 is arranged in the vehicle.

As shown in FIG. 2, the control / processing device 50 includes an analysis control unit 51, a mass spectrometry data processing unit 52, a storage unit 53, a learning data acquisition unit 54, a prediction model generation unit 55, and a sweet spot. It includes a prediction unit 56 and a display control unit 57. The substance of the control / processing device 50 is a personal computer provided with a large-capacity storage device such as a CPU, a memory, and a hard disk device, or a computer having a higher performance than that, and is stored in advance in a storage unit 53 composed of the large-capacity storage device. By executing the stored control / processing program on the CPU, the analysis control unit 51, the mass analysis data processing unit 52, the learning data acquisition unit 54, the prediction model generation unit 55, the sweet spot prediction unit 56, and the display control The function of the unit 57 is realized. The storage unit 53 includes a data storage unit 58 that stores data acquired by the mass spectrometer 10 and data related to sweet spot prediction.

The control / processing device 50 is connected to an input unit 60 including a pointing device such as a mouse and a keyboard, and a display unit 70 including a display device such as a liquid crystal display.

The analysis control unit 51 controls each unit of the mass spectrometer 10 according to the program. The detection signal by the ion detector 33 of the mass spectrometer 10 is converted into a digital value and then input to the mass spectrometry data processing unit 52. The details of the processing in the mass spectrometry data processing unit 52, and the functions of the data storage unit 58, the learning data acquisition unit 54, the prediction model generation unit 55, the sweet spot prediction unit 56, and the display control unit 57 will be described later.

Hereinafter, in the mass spectrometry system according to the present embodiment, a procedure for generating a prediction model for predicting a sweet spot will be described. Here, the “sweet spot” means a region in which the ionic strength obtained by MALDI mass spectrometry is larger than a predetermined threshold value. The "ionic strength" may be a relative value or an absolute value.

First, the user (analyzer) drops a mixed solution of the learning sample and the matrix into each well of the sample plate 21, and the mixture is dried and crystallized. The region on the sample plate 21 to which the mixed dried product of the learning sample and the matrix is attached corresponds to the "adhesion region of a predetermined sample" in the present invention. Here, as the learning sample, for example, a standard sample relating to a compound of the same type as the compound to be analyzed in an unknown sample described later is used. For example, when the unknown sample is a peptide, a peptide standard sample is used as the learning sample, when the unknown sample is sugar, a sugar standard sample is used as the learning sample, and the unknown sample is used. In the case of glycoprotein, a standard sample of glycoprotein is used as the learning sample. Further, as the matrix, it is desirable to use a matrix that is frequently used in the MALDI method and in which the localization of sample components is likely to occur in the process of crystallization. Examples of such a matrix include DHB (2,5-Dihydroxybenzoic Acid: 2,5-dihydroxybenzoic acid), DAN (1,5-Diaminonaphthalene: 1,5-diaminonaphthalene), or SA (Sinapinic Acid: sinapinic acid). Acid) and the like can be used.

Subsequently, the user sets the sample plate 21 on the sample stage 22 of the mass spectrometer 10 and then performs a predetermined operation via the input unit 60 to instruct the user to take a sample image and execute the mass analysis.

As a result, the stage drive unit 23 moves the sample stage 22 under the control of the analysis control unit 51, so that a predetermined well on the sample plate is directly under the aperture provided on the aperture plate 24 (that is, irradiation of laser light). Position), and further, the imaging unit 40 photographs the region to which the sample S (here, the mixed dried product of the learning sample and the matrix) in the well is attached via the aperture. The optical image thus obtained (hereinafter referred to as a learning sample image) is sent to the control / processing device 50 and stored in the data storage unit 58. FIG. 4 shows an example of a sample image for learning.

Subsequently, raster measurement by the mass spectrometer 10 is performed on the predetermined well. In this raster measurement, mass spectrometry is sequentially performed for each of a plurality of measurement points preset in one well. In the mass spectrometry at each measurement point, the emission of the laser light by the laser emitting unit 26 and the analysis by the mass spectrometry unit 30 are repeatedly performed a predetermined number of times (for example, several tens to several hundreds).

The detection signal from the ion detector 33 in the mass spectrometry is digitized and then sent to the mass spectrometry data processing unit 52 of the control / processing apparatus 50. The mass spectrometry data processing unit 52 integrates the detection signals obtained by a plurality of mass spectrometrys at each measurement point as described above to obtain the relationship between the m / z of ions generated from the measurement points and the intensity. Generate the mass spectrometric data shown. Further, in the mass spectrometry data processing unit 52, a specific m / z specified in advance by the user based on the two-dimensional coordinates representing the position of each measurement point in the well and the mass spectrum data for each measurement point. Alternatively, an image showing a two-dimensional distribution of signal intensity (that is, ion intensity) in a specific m / z range (hereinafter, collectively referred to as “objective m / z”) is created. This image is a so-called heatmap showing the difference in signal intensity (that is, ionic strength) with respect to the target m / z ion by the difference in hue, saturation, or lightness. Hereinafter, this image is referred to as a heat map image. The heat map image may be a color image in which the difference in signal strength is represented by a difference in color, or may be a monochrome image in which the difference in signal strength is represented by a difference in shade of black and white. FIG. 5 shows an example of a heat map image. The correspondence between the signal strength and the hue, saturation, or lightness, or the correspondence between the signal strength and the shade of black and white is stored in advance in the data storage unit 58. The heat map image data generated as described above is stored in the data storage unit 58 in association with the learning sample image relating to the same well. This heat map image corresponds to the "information indicating the position of the sweet spot" in the present invention. A region having a specific color (a predetermined color as a color indicating a region having a high ionic strength of the target m / z) on the heat map image corresponds to a sweet spot.

When the acquisition of the sample image for learning, the raster measurement, and the generation of the heat map image based on the result of the raster measurement are completed for one well, the stage drive unit 23 is under the control of the analysis control unit 51. By moving the sample stage 22, another well on the sample plate 21 is placed directly below the aperture provided on the aperture plate 24. Then, the above steps are repeatedly executed for all the wells designated by the user in advance until the acquisition of the learning sample image and the generation of the heat map image are completed. Hereinafter, the learning sample image and the heat map image associated therewith are referred to as learning data.

After that, when the user inputs a predetermined instruction from the input unit 60, the control / processing device 50 generates a prediction model for predicting the sweet spot. Hereinafter, the prediction model generation process will be described with reference to the flowchart of FIG.

In the control / processing device 50 that has received the instruction, first, the learning data acquisition unit 54 reads out a plurality of the learning data stored in the data storage unit 58 (step S11). Subsequently, the prediction model generation unit 55 generates a prediction model by a predetermined machine learning method using the plurality of learning data (step S12), and stores the prediction model in the data storage unit 58 (step). S13). The prediction model in the present embodiment is obtained, for example, when raster measurement by MALDI mass analysis is performed on the unknown sample from the brightness distribution in the optical image of the unknown sample provided in the well. This is an algorithm for generating a predicted heat map image (hereinafter referred to as a predicted heat map image). The position of the sweet spot depends on the dry state of the sample (that is, the shape, size, distribution, etc. of the mixed crystal grains of the sample and the matrix), and the brightness distribution in the optical image is the dry state. It reflects.

The machine learning method used to generate the prediction model is not particularly limited as long as it performs supervised learning, but for example, neural networks, deep learning, GAN (Generative Adversarial Networks), support vector machines, and the like. A random forest or the like can be used.

Subsequently, a method of predicting the sweet spot in the unknown sample formed on the well using the prediction model generated above will be described with reference to the flowchart of FIG.

First, the user drops a mixture of an unknown sample and a matrix into a predetermined well of the sample plate 21, which is dried and crystallized. At this time, it is desirable that the sample plate 21 is the same as or the same as the sample plate 21 used at the time of mass spectrometry of the learning sample (that is, at the time of collecting the learning data). Further, as the matrix, the same type of matrix as that used at the time of mass spectrometry of the learning sample is used.

Subsequently, the user sets the sample plate 21 on the sample stage 22 of the mass spectrometer 10 and then performs a predetermined operation on the input unit 60 to instruct the execution of the sweet spot prediction. As a result, the stage drive unit 23 moves the sample stage 22 under the control of the analysis control unit 51 to position the predetermined well directly under the aperture provided on the aperture plate 24, and further, the photographing unit 40 moves the sample stage 22. The region to which the sample S (here, the mixed dried product of the unknown sample and the matrix) in the predetermined well is attached is photographed. The optical image thus obtained (hereinafter referred to as an unknown sample image) is sent to the control / processing device 50 and temporarily stored in the data storage unit 58.

Subsequently, the sweet spot prediction unit 56 reads the unknown sample image and the above-mentioned prediction model from the data storage unit 58 (step S21), and inputs the unknown sample image into the prediction model to obtain the unknown sample. A predicted heat map image is generated (step S22). A region to which a specific color (a predetermined color indicating a region having high signal intensity) is attached on this predicted heat map image is a position predicted as a sweet spot (hereinafter referred to as a predicted sweet spot position). Is. That is, step S22 corresponds to "prediction of sweet spot position" in the present invention.

The display control unit 57 displays the predicted heat map image generated as described above on the screen of the display unit 70 (step S23), and displays the region to be analyzed by the user (hereinafter referred to as the analysis target region) as the predicted heat map image. It is specified above (step S24). At this time, the user designates an appropriate area from the areas colored with the specific color in the predicted heat map image as an analysis target area by clicking or tapping with a pointing device such as a mouse. The predicted heat map image may be displayed independently on the display unit 70, but the predicted heat map image may be displayed side by side with the unknown sample image, or the predicted heat map image may be superimposed on the unknown sample image. You may do it.

When the analysis target area is designated as described above, the analysis control unit 51 controls each part of the mass spectrometer 10 to perform laser irradiation and mass spectrometry on the analysis target area a plurality of times (for example, tens to hundreds of times). ) To be executed (step S25). Then, mass spectrometric data is generated by integrating the detection signals output from the ion detector 33 at each of the plurality of mass spectrometric analyzes by the mass spectrometric data processing unit 52, and the mass spectrometric data is the unknown. It is stored in the data storage unit 58 as the final mass spectrometric result of the sample.

As described above, according to the MALDI mass spectrometry system according to the present embodiment, a prediction model is generated using the training data collected by mass spectrometry of the training sample in advance, and an unknown sample image is added to the prediction model. By applying, the position of the sweet spot in the adhesion region of the unknown sample can be predicted. Therefore, as in the past, raster measurement is performed on an unknown sample, or the sweet spot in the adhesion region of the unknown sample is specified in advance by preliminary measurement, and then actual measurement (measurement to obtain the final mass spectrometry result) is performed. There is no need to do it, and the throughput of analysis can be improved. Further, as described above, when the MALDI mass analysis is performed on the training sample, the number of laser irradiations is increased by performing the laser irradiation as many times as the actual measurement of the unknown sample (that is, several tens to several hundreds). Finding a measurement point where the ion intensity increases with repeated laser irradiation, which could not be found by a small number of preliminary measurements, is also found as a sweet spot (that is, the measurement point is predicted to be a sweet spot). Can be done.

Although the embodiments for carrying out the present invention have been described above, the present invention is not limited to the above-described embodiments, and modifications are permitted within the scope of the gist of the present invention.

For example, in the above embodiment, the prediction result (that is, the prediction heat map image) by the prediction model is displayed on the screen of the display unit 70, and the analysis target area is specified by the user based on the prediction result. However, the control / processing device 50 may automatically determine the analysis target area based on the prediction result and perform MALDI mass spectrometry on the analysis target area. The flow chart of FIG. 7 shows the flow of processing in the control / processing device 50 in this case. Also in this case, the processes from reading the unknown sample image from the data storage unit to generating the predicted heat map image (that is, steps S31 to S32 in FIG. 7) are described in steps S21 to S22 in FIG. Since they are the same, the description thereof will be omitted here. In this example, after the predicted heat map image is generated, the control / processing device 50 determines one or more positions from the predicted sweet spot positions on the predicted heat map image based on a predetermined reference. It is selected as the analysis target area (step S33). After that, the analysis control unit 51 causes the mass spectrometer 10 to perform MALDI mass spectrometry on the analysis target region (step S34).

Further, in the above embodiment, the predicted heat map image is output as the prediction result of the sweet spot position by the prediction model, but the prediction model in the present invention is not limited to this, for example, the sweet spot. As a result of predicting the position, the signal strength of the target m / z is a predetermined threshold value on the unknown sample image (or the figure showing the contour of the sample area generated from the unknown sample image or the figure showing the contour of the well). An image with a mark indicating a position predicted to be the above may be output. Alternatively, the prediction model in the present invention has two-dimensional coordinates (that is, the above-mentioned X-axis and Y) of the position where the signal strength of the target m / z is predicted to be equal to or higher than a predetermined threshold as the prediction result of the sweet spot position. It may output (coordinates in the plane represented by the axis).

Further, in the above embodiment, machine learning is performed using a set of an optical image of a learning sample and a heat map image obtained by rasterly measuring the learning sample as learning data. The data for use is not limited to this. For example, a combination of the optical image of the learning sample and the mass spectrum data of each measurement point obtained by the raster measurement of the learning sample may be used as the learning data. Alternatively, in the optical image of the training sample and the raster measurement of the training sample, the position where the signal strength (relative value or absolute value) of the target m / z is equal to or higher than a predetermined threshold value (that is, the position of the sweet spot). ) May be used as learning data. The data representing the position where the signal strength of the target m / z is equal to or higher than a predetermined threshold value includes, for example, a learning sample image (or a figure showing the outline of a sample area generated from the learning sample image). Or, a figure showing the outline of the well) with a mark indicating a position where the signal strength of the target m / z is equal to or higher than a predetermined threshold value can be used. Alternatively, the two-dimensional coordinates of the position can be used as the data representing the position where the signal strength of the target m / z is equal to or higher than a predetermined threshold value.

Further, in the above embodiment, the learning sample image and the heat map image are newly collected by performing MALDI mass spectrometry of the training sample in order to generate the prediction model. Instead of this, the data storage unit The combination of the optical image of the sample taken by the past MALDI mass spectrometer stored in 58 and the raster measurement result of the sample may be used as the learning data.

In addition, since a skilled analyst can identify the sweet spot with relatively high accuracy based on the shape, size, distribution, etc. of the crystal grains in the sample image, past MALDI mass spectrometry The sample image taken at the time of , Can also be used as learning data in the present invention. In addition, a plurality of optical images of the sample are prepared, and a skilled analyst is made to select a position considered to be a sweet spot in each optical image, and information on each optical image and a position determined to be a sweet spot for the optical image. It is also possible to use the data associated with the above as the learning data in the present invention.

[Various aspects]
It will be understood by those skilled in the art that the above-described exemplary embodiments are specific examples of the following embodiments.

(Section 1) The method for predicting sweet spots according to one aspect is
Learning data in which the optical image obtained by photographing the adhesion region of the predetermined sample on the sample plate and the information indicating the position of the sweet spot in the adhesion region of the predetermined sample are associated with each other is acquired.
A prediction model that predicts the position of the sweet spot is generated by machine learning using the training data.
By applying the prediction model to an optical image obtained by photographing an adhesion region of an unknown sample, the position of a sweet spot in the adhesion region of the unknown sample is predicted.

(Section 2) The method for predicting sweet spots described in paragraph 1 is as follows.
The information indicating the position of the sweet spot included in the training data is mass spectrometry data obtained by performing MALDI mass spectrometry for each of a plurality of measurement points set in the adhesion region of the predetermined sample. There may be.

(Section 3) The method for predicting sweet spots described in paragraph 2 is as follows.
The mass spectrometric data may be a heat map image showing a two-dimensional distribution of ionic strength in the adhesion region of the predetermined sample.

(Section 4) The method for predicting sweet spots described in paragraph 1 is as follows.
The optical image included in the training data is an optical image taken during past MALDI mass spectrometry.
The information indicating the position of the sweet spot associated with the optical image may be the information of the analysis target position designated by the person in charge of analysis at the time of the past MALDI mass spectrometry.

(Section 5) The sweet spot prediction device according to one aspect is
A learning data acquisition unit that acquires learning data in which an optical image obtained by photographing an adhesion region of a predetermined sample on a sample plate and information indicating the position of a sweet spot in the adhesion region of the predetermined sample are associated with each other. When,
A prediction model generation unit that generates a prediction model that predicts the position of the sweet spot by executing machine learning using the training data.
A prediction unit that predicts the position of the sweet spot in the adhesion region of the unknown sample by applying the prediction model to the optical image obtained by photographing the adhesion region of the unknown sample.
It has.

(Section 6) The sweet spot prediction device according to paragraph 5 is
The information indicating the position of the sweet spot included in the training data is mass spectrometry data obtained by performing MALDI mass spectrometry for each of a plurality of measurement points set in the adhesion region of the predetermined sample. There may be.

(Section 7) The sweet spot prediction device according to paragraph 6 is
The mass spectrometric data may be a heat map image showing a two-dimensional distribution of ionic strength in the adhesion region of the predetermined sample.

(Item 8) The sweet spot prediction device according to item 5 is
The optical image included in the training data is an optical image taken during past MALDI mass spectrometry.
The information indicating the position of the sweet spot associated with the optical image may be the information of the analysis target position designated by the person in charge of analysis at the time of the past MALDI mass spectrometry.

(Section 9) The program according to one aspect causes a computer to function as each part of the sweet spot prediction device according to any one of paragraphs 5 to 8.

(Section 10) The non-transitory computer-readable medium according to one aspect stores the program described in paragraph 9.

According to the sweet spot prediction method according to paragraph 1, the sweet spot predictor according to paragraph 5, the program according to paragraph 9, or the computer-readable medium according to paragraph 10, the sweet in MALDI mass spectrometry. The position of the spot can be easily predicted.

Further, according to the sweet spot prediction method according to the second item or the sweet spot prediction device according to the sixth item, the mass spectrometry data obtained by actually executing the MALDI mass spectrometry is used as the training data. As a result, the sweet spot can be predicted with high accuracy.

Further, according to the sweet spot prediction method according to the third item or the sweet spot prediction device according to the seventh item, by using the heat map image as the mass spectrometry result, for example, each measurement is performed as the mass spectrometry result. Compared with the case of using a set of mass spectrometric data at points, the load on the computer in generating the prediction model can be reduced.

Further, according to the sweet spot prediction method according to the fourth item or the sweet spot prediction device according to the eighth item, it is not necessary to perform a new MALDI mass spectrometry for generating the prediction model. It is possible to suppress the work load of the user related to the generation.

10 ... Mass spectrometer 20 ... Ion source 26 ... Laser emitting unit 30 ... Mass spectrometer 33 ... Ion detector 40 ... Imaging unit 50 ... Control / processing device 51 ... Analysis control unit 52 ... Mass spectrometry data processing unit 54 ... For learning Data acquisition unit 55 ... Prediction model generation unit 56 ... Sweet spot prediction unit 57 ... Display control unit 58 ... Data storage unit 60 ... Input unit 70 ... Display unit

Claims (9)

Learning data in which the optical image obtained by photographing the adhesion region of the predetermined sample on the sample plate and the information indicating the position of the sweet spot in the adhesion region of the predetermined sample are associated with each other is acquired.
A prediction model that predicts the position of the sweet spot is generated by machine learning using the training data.
By applying the prediction model to the optical image obtained by photographing the adhesion region of the unknown sample, the position of the sweet spot in the adhesion region of the unknown sample is predicted.
How to predict sweet spots. The information indicating the position of the sweet spot included in the training data is mass spectrometry data obtained by performing MALDI mass spectrometry for each of a plurality of measurement points set in the adhesion region of the predetermined sample. The method for predicting a sweet spot according to claim 1. The method for predicting sweet spots according to claim 2, wherein the mass spectrometric data is a heat map image showing a two-dimensional distribution of ionic strength in the adhesion region of the predetermined sample. The optical image included in the training data is an optical image taken during past MALDI mass spectrometry.
The sweet spot according to claim 1, wherein the information indicating the position of the sweet spot associated with the optical image is the information of the analysis target position designated by the analyst at the time of the past MALDI mass spectrometry. Prediction method. A learning data acquisition unit that acquires learning data in which an optical image obtained by photographing an adhesion region of a predetermined sample on a sample plate and information indicating the position of a sweet spot in the adhesion region of the predetermined sample are associated with each other. When,
A prediction model generation unit that generates a prediction model that predicts the position of the sweet spot by executing machine learning using the training data.
A prediction unit that predicts the position of the sweet spot in the adhesion region of the unknown sample by applying the prediction model to the optical image obtained by photographing the adhesion region of the unknown sample.
Sweet spot predictor with. The information indicating the position of the sweet spot included in the training data is mass spectrometry data obtained by performing MALDI mass spectrometry for each of a plurality of measurement points set in the adhesion region of the predetermined sample. The sweet spot prediction device according to claim 5. The sweet spot prediction device according to claim 6, wherein the mass spectrometric data is a heat map image showing a two-dimensional distribution of ionic strength in the adhesion region of the predetermined sample. The optical image included in the training data is an optical image taken during past MALDI mass spectrometry.
The sweet spot prediction according to claim 5, wherein the information indicating the position of the sweet spot associated with the optical image is the information of the analysis target position designated by the analyst at the time of the past MALDI mass spectrometry. Device. A program that causes a computer to function as each part of the sweet spot predictor according to any one of claims 5 to 8.

Images