Information processing device, method for operating information processing device, and program for operating information processing device

Abstract

An information processing device that predicts the state of a target component in a suspension in which biomolecules are dispersed as components in a liquid on the basis of spectral data obtained by measuring electromagnetic waves emitted from the suspension, the information processing device being provided with a processor, the processor performs: a process using a state prediction model corrected from reference spectral data obtained from a reference suspension in a reference measurement environment; a target spectroscopic data acquisition unit that acquires target spectroscopic data obtained from a target suspension in which the state of a target component is unknown in a target measurement environment different from the reference measurement environment; acquiring a correction coefficient derived by comparing the reference spectroscopic data corresponding to the target spectroscopic data and reference spectroscopic data obtained from the reference suspension in the target measurement environment; correcting the target spectral data according to the correction coefficient to generate corrected target spectral data; and causing the state prediction model to predict the state of the object component by applying the correction object spectral data to the state prediction model.

Description

TECHNICAL FIELD

[0001] The technology of the present application relates to an information processing apparatus, a method of operation of an information processing apparatus, and a program of operation of an information processing apparatus. BACKGROUND

[0002] For example, a manufacturing process of a bio-pharmaceutical product using a biological molecule such as a protein or the like as an effective component is known. In such a manufacturing process, a suspension in which various components represented by the effective component are dispersed in a liquid is often produced. Monitoring the state of a target component (for example, a protein or a protein aggregate) among the components in the suspension (for example, the concentration of the protein or the concentration of the protein aggregate) is important for successfully guiding the ongoing manufacturing process.

[0003] As a technique for predicting the state of the target component, a technique described below is attracting attention from the reasons that there are few concerns about contamination and the like. That is, a technique in which electromagnetic waves such as Raman scattered light emitted from the suspension are measured, and spectral data such as Raman spectrum data obtained thereby are applied to a state prediction model such as a multivariate analysis or a machine learning model, whereby the state prediction model predicts the state of the target component.

[0004] In addition, a technique for correcting the deviation of Raman spectrum data caused by the intensity variation of excitation light from a light source is described in Japanese Patent Application Publication No. 09-089775. Also, a technique for suppressing the deviation of a plurality of Raman spectrum data by performing derivative transformation and standard normal variate (SNV) on the plurality of Raman spectrum data is described in Japanese Patent Application Publication No. 2022-552876. SUMMARY

[0005] Technical Problem to be Solved by the Invention In the manufacturing process of a bio-pharmaceutical product, condition setting is initially performed using a relatively small-scale device, and after the end of the condition setting, the scale of the device being processed is gradually expanded by moving to a relatively large-scale device or the like. With the expansion of the scale of the device, for example, the measurement instrument for stably measuring electromagnetic waves, that is, a flow cell or the like, is changed, and the measurement environment of the electromagnetic waves also changes in various ways.

[0006] If the measurement environment of the electromagnetic waves changes, sometimes the spectral data will be different even for a suspension of the same composition. Therefore, the prediction result of the state of the target component of the suspension of the same composition should be the same, but since the spectral data originating from the measurement environment is different, different prediction results will be brought about. In short, it is possible that the reliability of the prediction result of the state of the target component will be reduced due to the change in the measurement environment of the electromagnetic waves.

[0007] One embodiment of the technology of the present application provides an information processing apparatus capable of obtaining a prediction result of a state of a target component in a suspension liquid with high reliability even when a measurement environment of an electromagnetic wave changes, a method of operation of the information processing apparatus, and a program of operation of the information processing apparatus.

[0008] Means for solving technical problems The information processing apparatus of the present application predicts a state of a target component in a suspension liquid from a spectrum obtained by measuring an electromagnetic wave emitted from a suspension liquid in which a biomolecule is dispersed in a liquid, the information processing apparatus comprising a processor that performs processing of: using a state prediction model corrected from reference spectrum data obtained from a reference suspension liquid in a reference measurement environment; acquiring object spectrum data obtained from an object suspension liquid in which a state of a target component is unknown in an object measurement environment different from the reference measurement environment; acquiring a correction coefficient derived from a comparison of the reference spectrum data and reference spectrum data corresponding to the object spectrum data and obtained from the reference suspension liquid in the object measurement environment; generating corrected object spectrum data by correcting the object spectrum data according to the correction coefficient; and causing the state prediction model to predict the state of the target component by applying the corrected object spectrum data to the state prediction model.

[0009] It is preferable that the correction coefficient be derived from a ratio of intensity values of the reference spectrum data and the reference spectrum data corresponding to a component in the reference suspension liquid.

[0010] It is preferable that the correction coefficient be derived after removing a ratio of intensity values corresponding to a measurement instrument of the electromagnetic wave.

[0011] It is preferable that the measurement instrument be a flow cell having a flow path for a suspension liquid to flow and an optical system for taking in an electromagnetic wave, and the intensity value corresponding to the measurement instrument be an intensity value corresponding to the optical system.

[0012] It is preferable that the reference measurement environment and the object measurement environment differ in any one of presence or absence of use and a kind of a measurement instrument of the electromagnetic wave.

[0013] It is preferable that the measurement instrument be a flow cell having a flow path for a suspension liquid to flow and an optical system for taking in an electromagnetic wave.

[0014] It is preferable that there be a plurality of object measurement environments that differ in a kind of a flow cell used.

[0015] It is preferable that the correction coefficient be stored in advance in a storage section for each of the plurality of object measurement environments, and the processor performs processing of: receiving a designation of a kind of a flow cell used; and acquiring the correction coefficient corresponding to the designation by reading from the storage section.

[0016] It is preferable that at least any one of the distance between the incident surface of the electromagnetic wave of the optical system of the flow cell and the wall surface of the flow path opposite to the incident surface, the reflectivity of the wall surface, and the focal length of the optical system be different.

[0017] It is preferable that the state prediction model be corrected based on the calibration target spectrum data.

[0018] It is preferable that the state of the target component be the concentration of the target component.

[0019] It is preferable that the target component be a protein.

[0020] It is preferable that the protein be an antibody.

[0021] It is preferable that the electromagnetic wave be Raman scattered light.

[0022] It is preferable that the state prediction model be a machine learning model that learns using reference spectrum data as teaching data.

[0023] The operation method of the information processing apparatus of the present application is an operation method of an information processing apparatus that predicts the state of a target component in a suspension in which a biomolecule is dispersed in a liquid based on spectrum data obtained by measuring electromagnetic waves emitted from the suspension, and includes the steps of: using a state prediction model corrected based on reference spectrum data obtained from a reference suspension in a reference measurement environment; acquiring target spectrum data obtained from a target suspension in which the state of the target component is unknown in a target measurement environment different from the reference measurement environment; acquiring a correction coefficient derived by comparing the reference spectrum data and reference spectrum data corresponding to the target spectrum data and obtained from the reference suspension in the target measurement environment; generating calibration target spectrum data by correcting the target spectrum data based on the correction coefficient; and causing the state prediction model to predict the state of the target component by applying the calibration target spectrum data to the state prediction model.

[0024] The operation program of the information processing apparatus of the present application is an operation program of an information processing apparatus that predicts the state of a target component in a suspension in which a biomolecule is dispersed in a liquid based on spectrum data obtained by measuring electromagnetic waves emitted from the suspension, and causes a computer to execute processing including the steps of: using a state prediction model corrected based on reference spectrum data obtained from a reference suspension in a reference measurement environment; acquiring target spectrum data obtained from a target suspension in which the state of the target component is unknown in a target measurement environment different from the reference measurement environment; acquiring a correction coefficient derived by comparing the reference spectrum data and reference spectrum data corresponding to the target spectrum data and obtained from the reference suspension in the target measurement environment; generating calibration target spectrum data by correcting the target spectrum data based on the correction coefficient; and causing the state prediction model to predict the state of the target component by applying the calibration target spectrum data to the state prediction model.

[0025] Inventive Effects According to the technology of the present application, it is possible to provide an information processing apparatus capable of obtaining a prediction result of a state of a target component in a suspension liquid with high reliability even when a measurement environment of electromagnetic waves changes, a method of operating an information processing apparatus, and a program for operating an information processing apparatus. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a diagram showing a concentration unit, a measurement system, and an information processing apparatus.

[0027] Figure 2 is a diagram showing excitation light and Raman scattered light.

[0028] Figure 3 is a diagram showing Raman spectral data.

[0029] Figure 4 is a cross-sectional view of a flow cell.

[0030] Figure 5 is a block diagram of a computer constituting an information processing apparatus.

[0031] Figure 6 is a block diagram of a processing unit of a CPU of an information processing apparatus.

[0032] Figure 7 is a diagram showing correction coefficient information.

[0033] Figure 8 is a diagram showing a case of measurement of reference Raman spectral data and reference Raman spectral data.

[0034] Figure 9 is a table showing contents of a reference suspension liquid.

[0035] Figure 10 is a diagram showing a correction coefficient derivation process.

[0036] Figure 11 is a diagram showing a flow cell selection screen.

[0037] Figure 12 is a diagram showing a case where a correction coefficient corresponding to a designation of a kind of a flow cell used is read and acquired from a storage device.

[0038] Figure 13 is a diagram showing a process of a correction unit.

[0039] Figure 14 is a diagram showing a process of a prediction unit.

[0040] Figure 15 is a diagram showing a neural network constituting an antibody concentration prediction model.

[0041] Figure 16 This is a graph representing the processing during the learning phase of the antibody concentration prediction model.

[0042] Figure 17 This is a diagram representing the screen displaying the prediction results.

[0043] Figure 18 It is a flowchart illustrating the processing steps of an information processing device.

[0044] Figure 19 This is another example of a graph representing information about correction coefficients.

[0045] Figure 20 This is another example of a graph representing correction coefficient information.

[0046] Figure 21 This is a diagram illustrating the second implementation of an antibody concentration prediction model learned from the Raman spectral data of the calibration object.

[0047] Figure 22 This is a graph showing the predicted and measured values ​​of antibody concentrations in the examples, and their average error.

[0048] Figure 23 This is a table representing the contents of the comparison examples.

[0049] Figure 24 This is a graph showing the predicted and measured values ​​of antibody concentrations in Comparative Example 1, and their average error.

[0050] Figure 25 This is a graph showing the predicted and measured antibody concentrations of Comparative Example 2 and their average error.

[0051] Figure 26 This is a graph showing the predicted and measured values ​​of antibody concentrations in Comparative Example 3, and their average error. Detailed Implementation

[0052] [First Implementation] As an example, such as Figure 1 As shown, the assay system 10 includes a flow cell 11 and a Raman spectrometer 12. The assay system 10 is, for example, assembled in the concentration section 15 after the culture section 13 and the purification section 14 in a biopharmaceutical active pharmaceutical ingredient manufacturing system.

[0053] The cultivation section 13 has a cultivation tank and a cell removal filter. The cell cultivation solution (culture medium) is stored in the cultivation tank. The antibody-producing cells are seeded in the cell cultivation solution, and the antibody-producing cells are cultivated in the cell cultivation solution. The cultivation method can be any one of perfusion cultivation and fed-batch cultivation. The antibody-producing cells are, for example, cells established by integrating an antibody gene into a host cell such as a Chinese hamster ovary cell (CHO). The antibody-producing cells produce immunoglobulin, i.e., antibody 16, during cultivation. Therefore, the antibody-producing cells and the antibody 16 are present in the cell cultivation solution. The antibody 16 is, for example, a monoclonal antibody, and is an effective ingredient of a biological medicine. The antibody 16 is an example of the "ingredient" and the "target ingredient" and the "protein" related to the technology of the present application.

[0054] The cell removal filter captures the antibody-producing cells in the cell cultivation solution with a filter membrane by, for example, a tangential flow filtration (TFF) method or an alternating tangential flow filtration (ATF) method, and removes the antibody-producing cells from the cell cultivation solution. Also, the cell removal filter allows the antibody 16 to pass through. Therefore, the cell cultivation solution from the cultivation section 13 to the purification section 14 mainly contains the antibody 16. In this way, the cell cultivation solution obtained by removing the antibody-producing cells with the cell removal filter is called a cultivation supernatant. In addition, the cultivation supernatant contains, in addition to the antibody 16, cell-derived proteins, cell-derived DNA (Deoxyribonucleic Acid), aggregates 17 of the antibody 16, or viruses, and the like. These cell-derived proteins, cell-derived DNA, aggregates 17 of the antibody 16, and viruses are also examples of the "ingredient" related to the technology of the present application.

[0055] The purification section 14 removes impurities and viruses in stages and increases the purity of the antibody 16 in the cultivation supernatant in stages by sequentially performing ingredient separation treatment based on a plurality of chromatography devices on the cultivation supernatant from the cultivation section 13. The plurality of chromatography devices are, for example, an immunoaffinity chromatography device, a size exclusion chromatography device, a cation chromatography device, an anion chromatography device, a hydrophobic interaction chromatography device, and the like.

[0056] Further, the purification section 14 performs a process of inactivating viruses in the culture supernatant and a process of removing viruses in the culture supernatant by a filter. The purified liquid 18 obtained through such various processes is sent from the purification section 14 to the concentration section 15. In addition, a filter of a single pass tangential flow filtration (SPTFF) system can be provided in front of various chromatography devices.

[0057] A primary tank 20, a pump 21, a concentration filter 22, a secondary tank 23, a pressure control valve 24, and the like are provided in the concentration section 15. The flow-in path 25, the flow-out path 26, and the return flow path 27 are connected to the primary tank 20. The purified liquid 18 from the purification section 14 flows into the primary tank 20 through the flow-in path 25. The primary tank 20 stores the purified liquid 18.

[0058] The pump 21 is provided in the flow-out path 26. By driving the pump 21, the purified liquid 18 stored in the primary tank 20 is sent to the flow-out path 26 and introduced into the concentration filter 22 through the flow-out path 26. The concentration filter 22 performs, for example, concentration / filtration processes based on ultrafiltration (UF) and diafiltration (DF) on the introduced purified liquid 18.

[0059] By the concentration / filtration processes of the concentration filter 22, the purified liquid 18 is separated into a concentrated liquid in which the purity of the antibody 16 is further increased and a waste liquid 28 of water, a solvent, or the like. The concentrated liquid returns to the primary tank 20 through the return flow path 27. Thus, the liquid stored in the primary tank 20 becomes a mixed liquid 18M of the purified liquid 18 from the purification section 14 and the concentrated liquid from the concentration filter 22. The purified liquid 18 and the mixed liquid 18M are examples of the "suspension liquid" and the "target suspension liquid" to which the technology of the present application is directed. Hereinafter, unless specifically distinguished, the purified liquid 18 and the mixed liquid 18M are collectively referred to as the target suspension liquid 18T.

[0060] The waste liquid 28 is introduced into the secondary tank 23 through the waste liquid path 29. The secondary tank 23 stores the waste liquid 28.

[0061] The pressure control valve 24 is provided in the return flow path 27. By performing on-off control on this pressure control valve 24, the hydraulic pressure of the concentrated liquid passing through the return flow path 27 is kept constant. In addition, although not shown, a pressure gauge is provided in the downstream side of the pump 21 in the flow-out path 26, the upstream side of the pressure control valve 24 in the return flow path 27, and the waste liquid path 29. Further, the primary tank 20 and the secondary tank 23 are placed on a weight gauge. Based on the pressures of the respective liquids measured by the pressure gauges and the weights of the primary tank 20 and the secondary tank 23 measured by the weight gauge, the driving of the pump 21 and the pressure control valve 24 is controlled.

[0062] The flow tank 11 is connected to the upstream side of the pump 21 in the outflow path 26. In the flow tank 11, the object suspension 18T from the primary tank 20 flows at a preset flow rate.

[0063] As an example, such as Figure 2 As shown, the Raman spectrometer 12 is a device for evaluating a substance M by utilizing the characteristics of Raman scattered light RSL. When an excitation light EL is irradiated onto a substance M, Raman scattered light RSL with a different wavelength than the excitation light EL is generated through the interaction between the excitation light EL and the substance M. The wavelength difference between the excitation light EL and the Raman scattered light RSL corresponds to the energy portion of the molecular vibrations of the substance M. Therefore, Raman scattered light RSL with different wavenumbers can be obtained between substances M with different molecular structures. Raman scattered light RSL is an example of "electromagnetic wave" involved in the technology of this invention. In addition, the Raman scattered light RSL preferably uses Stokes lines and anti-Stokes lines.

[0064] The Raman spectrometer 12 consists of a sensor unit 35 and an analyzer 36. The sensor unit 35 is connected to the flow cell 11. The sensor unit 35 emits excitation light EL from its front end. The excitation light EL irradiates the target suspension 18T flowing in the flow cell 11. Raman scattered light RSL is generated through the interaction between the excitation light EL and components such as antibody 16 in the target suspension 18T. The sensor unit 35 receives the Raman scattered light RSL and outputs the received Raman scattered light RSL to the analyzer 36.

[0065] Analyzer 36 decomposes the Raman scattered light RSL at each wavenumber, deriving the intensity value of the Raman scattered light RSL at each wavenumber, thereby generating a sample as an example. Figure 3 The Raman spectral data 37, representing the Raman scattering light RSL, is shown below. Raman spectral data 37 records the intensity values ​​of the Raman scattering light RSL for each wavenumber. In this example, Raman spectral data 37 is measured at 1 cm⁻¹. -1 Scale-derived wavenumber 700cm -1 ~1800cm -1 Data on the intensity values ​​of Raman scattered light RSL within a certain range. Additionally, in Figure 3 In the diagram, the lower part of the Raman spectral data 37 shows a graph that plots the intensity values ​​of the Raman spectral data 37 by each wavenumber and connects them with lines. The Raman spectral data 37 is an example of the "spectral data" involved in the technology of this invention.

[0066] Thus, the measurement system 10 flows the target suspension 18T in the flow cell 11. Then, the sensor unit 35 irradiates the target suspension 18T flowing in the flow cell 11 with excitation light EL, thereby measuring the Raman scattered light RSL of components such as antibody 16 in the target suspension 18T, and obtaining Raman spectral data 37. Furthermore, Figure 1 The measurement environment shown, which uses Raman scattering light RSL in flow cell 11, is an example of the "object measurement environment" involved in the technology of this invention.

[0067] In the information processing device 40, a Raman spectrometer 12 is connected to a computer network such as a LAN (Local Area Network). The Raman spectrometer 12 transmits the measured Raman spectral data 37 to the information processing device 40. The Raman spectral data 37 transmitted from the Raman spectrometer 12 to the information processing device 40 is data used to predict the concentration of antibody 16. The concentration of antibody 16 is an example of the "state of the target component" according to the technology of this invention. In the following description, the Raman spectral data 37 transmitted from the Raman spectrometer 12 to the information processing device 40 will be labeled as target Raman spectral data 37T. Target Raman spectral data 37T is an example of "target spectral data" according to the technology of this invention.

[0068] The information processing device 40 is, for example, a desktop personal computer, equipped with a monitor 41 for displaying various screens and input devices 42 such as a keyboard, mouse, touch panel, and / or microphone for voice input. The information processing device 40 is, for example, installed in a pharmaceutical company developing biopharmaceuticals or an organization that receives biopharmaceutical development contracts from a pharmaceutical company, i.e., a Contract Research Organization (CRO). The information processing device 40 is operated by a user U who participates in the development of biopharmaceuticals at the pharmaceutical company or CRO (hereinafter collectively referred to as a pharmaceutical facility).

[0069] As an example, such as Figure 4 As shown, the flow cell 11 consists of a main body 50 and a sensor connector 51. The main body 50 is a cylindrical component with a straight flow path 52 with a circular cross-section at its center. The main body 50 is formed of a metal, such as Hastelloy. Alternatively, the main body 50 may be formed of a resin, such as a polyolefin resin.

[0070] A first connecting portion 53 and a second connecting portion 54 in the shape of a cylindrical boss are provided at the center of the both end surfaces of the main body 50. The first connecting portion 53 has a flow inlet 55 of the flow path 52, and the second connecting portion 54 has a flow outlet 56 of the flow path 52. The first connecting portion 53 and the second connecting portion 54 are, for example, parallel threads or tapered threads. A sterile connector provided in the flow-out path 26 is liquid-tightly attached to the first connecting portion 53 and the second connecting portion 54.

[0071] An attachment hole 57 for detachably attaching the sensor portion connector 51 to the main body 50 is formed at the center of the peripheral surface of the main body 50. A thread is formed on the inner peripheral surface of the attachment hole 57. The attachment hole 57 penetrates the flow path 52. A thread that screws with the thread of the inner peripheral surface of the attachment hole 57 is formed on the front end portion of the sensor portion connector 51. The sensor portion connector 51 is attached to the attachment hole 57 by screwing the threads with each other, and the front end portion of the sensor portion connector 51 is housed in the attachment hole 57. The front end of the sensor portion 35 is detachably connected to the sensor portion connector 51.

[0072] The sensor portion connector 51 is a cylindrical member in which an optical system 58 is provided at the front end. The optical system 58 is composed of a spherical lens 59 and a transparent plate 60. The optical axes of the spherical lens 59 and the transparent plate 60 coincide with each other. As the name implies, the spherical lens 59 is a lens in the shape of a sphere, and is made of, for example, sapphire glass or quartz glass. The spherical lens 59 condenses excitation light EL from the sensor portion 35, and inputs Raman scattered light RSL generated by the interaction of the excitation light EL and the antibody 16 or the like in the object suspension 18T into the sensor portion 35.

[0073] The transparent plate 60 is also made of, for example, sapphire glass or quartz glass like the spherical lens 59. The transparent plate 60 is a circular plate in which a first face 61 on the side of the spherical lens 59 is parallel to a second face 62 on the side of the flow path 52. The second face 62 is in contact with the object suspension 18T flowing in the flow path 52. The second face 62 is an example of the "incident face" according to the present technology. Here, "parallel" means parallel including an error within a range that is generally allowed in the technical field to which the present technology pertains and does not deviate from the gist of the present technology, in addition to complete parallel.

[0074] The condensing position of the spherical lens 59 is, for example, on the second face 62 of the transparent plate 60. The condensing position of the spherical lens 59 is determined by the diameter, the refractive index, the focal length FL, and the like of the spherical lens 59. The condensing position at this time is a point. By setting the condensing position of the spherical lens 59 to a position on the second face 62 of the transparent plate 60, it is possible to reduce the concern that the excitation light EL attenuates in the object suspension 18T. Furthermore, it is possible to maintain the S / N ratio of the Raman spectral data 37 at a high level.

[0075] The wall surface 63 forming the flow path 52 is, for example, a smooth surface without any unevenness greater than 1 mm. The wall surface 63A opposite to the second surface 62 of the transparent plate 60 has a reflectivity R. Furthermore, the distance D between the second surface 62 and the wall surface 63A is approximately the same as the diameter of the flow path 52. The diameter of the flow path 52 is set according to the flow rate, viscosity, and size of the thickening filter 22 of the object suspension 18T flowing in the flow path 52. For example, under the conditions that the flow rate of the object suspension 18T flowing in the flow path 52 is 200 cc / min or more and the viscosity of the object suspension 18T is 0.001 Pa·s, the same as that of water, the diameter of the flow path 52 is in the range where the Reynolds number Re in the flow path 52 is 2300 or more (Re≥2300). If the Reynolds number Re is 2300 or more, turbulence is generated in the object suspension 18T flowing in the flow path 52. As a result, the deviation of the components in the object suspension 18T is reduced, thereby improving the stability of the Raman spectral data 37.

[0076] As an example, such as Figure 5 As shown, in addition to the aforementioned display 41 and input device 42, the computer constituting the information processing device 40 also includes a storage device 70, a memory 71, a CPU (Central Processing Unit) 72, and a communication unit 73. They are interconnected via a bus 74.

[0077] Storage device 70 is either a computer built into the information processing device 40 or a hard disk drive connected via cable or network. Alternatively, storage device 70 may be a disk array that connects multiple hard disk drives. Storage device 70 stores control programs such as operating systems, various application programs, and various data associated with these programs. Storage device 70 is an example of a "storage unit" according to the technology of this invention. Alternatively, a solid-state drive can be used instead of a hard disk drive.

[0078] Memory 71 is the working memory used by CPU 72 to perform processing. CPU 72 loads the program stored in storage device 70 into memory 71 and executes the processing according to the program. Thus, CPU 72 centrally controls the various parts of the computer. CPU 72 is an example of a "processor" according to the technology of this invention. In addition, memory 71 may be built into CPU 72. Communication unit 73 performs various information transmission control with external devices such as Raman spectrometer 12.

[0079] As an example, such as Figure 6As shown, the storage device 70 stores a working program 80. The working program 80 is an application program used to enable the computer to function as an information processing device 40. In addition to the working program 80, the storage device 70 also stores correction coefficient information 81 and an antibody concentration prediction model 82. The antibody concentration prediction model 82 is an example of a "state prediction model" involved in the technology of this invention.

[0080] If the working program 80 is started, the CPU 72 and memory 71 of the computer constituting the information processing device 40 work together as the acquisition unit 85, the read / write (hereinafter referred to as RW) control unit 86, the instruction receiving unit 87, the correction unit 88, the prediction unit 89 and the display control unit 90.

[0081] The acquisition unit 85 acquires object Raman spectral data 37T from the Raman spectrometer 12. The acquisition unit 85 outputs the object Raman spectral data 37T to the RW control unit 86.

[0082] The RW control unit 86 controls the storage of various data to the storage device 70 and the retrieval of various data stored in the storage device 70. The RW control unit 86 stores the target Raman spectrum data 37T from the acquisition unit 85 into the storage device 70. Furthermore, the RW control unit 86 reads the target Raman spectrum data 37T and the correction coefficients CF (reference) registered in the correction coefficient information 81 from the storage device 70. Figure 7 The RW control unit 86 reads the antibody concentration prediction model 82 from the storage device 70 and outputs the read antibody concentration prediction model 82 to the prediction unit 89.

[0083] The instruction receiving unit 87 receives various operation instructions input by the user U via the input device 42.

[0084] The correction unit 88 generates the Raman spectrum data 37TC of the target object based on the correction coefficient CF. The correction unit 88 outputs the Raman spectrum data 37TC of the target object to the prediction unit 89.

[0085] The prediction unit 89 applies the Raman spectrum data 37TC of the calibration object to the antibody concentration prediction model 82, enabling the antibody concentration prediction model 82 to predict the concentration of antibody 16 in the suspension 18T, and outputs its prediction result 95 from the antibody concentration prediction model 82. The prediction unit 89 outputs the prediction result 95 to the display control unit 90.

[0086] The display control unit 90 controls the display of various screens on the display 41. For example, the display control unit 90 controls the display of the use flow pool selection screen 115 (see reference) for allowing the user U to select the flow pool 11 to be used. Figure 11 The prediction result display screen shows prediction result 95 (reference). Figure 17 (e.g., displayed on monitor 41.)

[0087] As an example, such as Figure 7 As shown, the correction coefficient information 81 is information that registers the correction coefficients CF for all types of flow cells 11 (product names FC1, FC2, FC3, ...) that can be used in the manufacturing system of biopharmaceuticals. At least one of the following is different: the distance D between the second surface 62 of the transparent plate 60 of the various flow cells 11 and the wall 63A of the flow path 52 opposite the second surface 62; the reflectivity R of the wall 63A; and the focal length FL of the spherical lens 59. Furthermore, the correction coefficient information 81 is not limited to the method of using the flow cell 11; it can also register correction coefficients CF for methods where the sensor unit 35 is directly immersed in a tank (chamber) containing the target suspension 18T without using the flow cell 11.

[0088] In biopharmaceutical manufacturing systems, for example, different flow cells 11 are selectively used depending on the scale of the equipment. Specifically, when setting conditions using a relatively small-scale device and then transferring to a relatively large-scale device after the condition setting is completed, the flow cell is changed for the large-scale device. Such selective use of multiple measurement environments with different flow cells 11 is an example of the "multiple object measurement environments" involved in the technology of this invention.

[0089] The correction factor CF is derived as follows. First, as an example, such as... Figure 8 As shown, a standard suspension of 18ST is prepared in two tanks, 100A and 100B. As an example, as... Figure 9 As shown in Table 105, there are five reference suspensions (18ST). Specifically, reference suspensions (18ST) include a 78 g / L solution of monoclonal antibody (mAb) as reference suspension 1, a 50 g / L solution of bovine serum albumin (BSA) as reference suspension 2, and a 25 g / L solution of bovine serum albumin as reference suspension 3. Furthermore, reference suspensions (18ST) include a 5 g / L solution of phenylalanine (Phe) as reference suspension 4 and a 5 g / L solution of tryptophan (Trp) as reference suspension 5.

[0090] The sensor section 35 of the Raman spectrometer 12 is directly immersed in the reference suspension 18ST within tank 100A, and the Raman scattered light RSL from the reference suspension 18ST is measured, thereby obtaining reference Raman spectral data 37ST. The measurement environment without using the Raman scattered light RSL from the flow cell 11 is an example of the "reference measurement environment" involved in the technology of this invention.

[0091] On the other hand, the reference suspension 18ST in tank 100B is flowed into flow cell 11, and sensor unit 35 is connected in flow cell 11 to measure the Raman scattered light RSL from the reference suspension 18ST flowing in flow cell 11, thereby obtaining reference Raman spectral data 37R. Reference Raman spectral data 37R is Raman spectral data corresponding to the target Raman spectral data 37T. Because the flow cell 11 is used in the measurement environment of this reference Raman spectral data 37R, it is consistent with... Figure 1 The measurement environment shown is also an example of the "object measurement environment" involved in the technology of this invention. That is, the reference measurement environment and the object measurement environment in this example differ in the presence or absence of the flow cell 11. In addition, the reference measurement environment and the object measurement environment can differ in the type of flow cell 11. That is, even in the reference measurement environment, a flow cell 11 with a pre-set specification can be used.

[0092] The obtained reference Raman spectral data 37ST and reference Raman spectral data 37R are processed by deriving correction coefficients 101.

[0093] As an example, such as Figure 10 As shown in Figure 110, the correction factor derivation process 101 begins by calculating the ratio (peak ratio) of the intensity value of the reference Raman spectral data 37ST corresponding to the component in reference suspension 18ST to the intensity value of the reference Raman spectral data 37R corresponding to the component in reference suspension 18ST for each reference suspension 1 to 5. If the component in reference suspension 18ST is reference suspension 1, it is a monoclonal antibody; if it is reference suspension 4, it is phenylalanine. For example, if the intensity value of the reference Raman spectral data 37ST corresponding to the component in reference suspension 18ST is 100, and the intensity value of the reference Raman spectral data 37R corresponding to the component in reference suspension 18ST is 95, their ratio is 100 / 95 ≈ 1.05. After calculating the ratio of intensity values ​​for each reference suspension 1 to 5 in this way, the average value of their intensity value ratios is calculated. This average value is used as the correction factor CF.

[0094] Here, when using flow cell 11, in the Raman spectral data 37, not only are peaks corresponding to the components in the target suspension 18T or reference suspension 18ST observed, but also peaks originating from the material of the spherical lens 59, i.e., sapphire glass, and other peaks originating from the optical system 58 of flow cell 11 are observed. These peaks originating from the optical system 58 are not observed in the reference measurement environment of this example without using flow cell 11, or their positions and / or intensity values ​​are significantly different. Therefore, as shown in Figure 111, the ratio of the intensity value of the reference Raman spectral data 37ST corresponding to the optical system 58 to the intensity value of the reference Raman spectral data 37R corresponding to the components in the reference suspension 18ST becomes a value that deviates from the ratio of the intensity value of the reference Raman spectral data 37ST corresponding to the components in the reference suspension 18ST to the intensity value of the reference Raman spectral data 37R corresponding to the components in the reference suspension 18ST. Therefore, in this example, as indicated by ×, the correction factor CF is derived after removing the ratio of the intensity values ​​corresponding to the optical system 58.

[0095] In the case where the instruction receiving unit 87 receives an instruction from user U to predict the concentration of antibody 16 based on object Raman spectral data 37T, as an example, the display control unit 90 performs... Figure 11 The flow cell selection screen 115 shown is displayed on the control panel of the monitor 41. The flow cell selection screen 115 includes a drop-down menu 116 for selecting the flow cell 11 to be used when the object Raman spectral data 37T is obtained. The drop-down menu 116 lists, in a selectable manner, characters that uniquely identify the flow cell 11, such as its product name and model number (product name is shown here). The user selects the flow cell 11 to be used using the drop-down menu 116 and then selects the OK button 117.

[0096] In the flow pool selection screen 115, if user U rotates the flow pool 11 used by the user through the drop-down menu 116 and selects the OK button 117, as an example, Figure 12 As shown, the instruction receiving unit 87 receives specified information 120 from the flow cell 11 used. The specified information 120 includes the product name, model, etc. of the flow cell 11 (product name is shown here). The instruction receiving unit 87 outputs the specified information 120 to the RW control unit 86.

[0097] The RW control unit 86 acquires the correction coefficient CF corresponding to the flow cell 11 specified by the designation information 120 by reading the correction coefficient information 81 from the storage device 70. The RW control unit 86 outputs the read correction coefficient CF to the correction unit 88. The correction unit 88 corrects the target Raman spectrum data 37T based on the correction coefficient CF from the RW control unit 86. Additionally, in Figure 12In the example shown, the flow cell 11 with the product name "FC1" is selected as the flow cell to be used, and "1.05" is read from the correction coefficient information 81 as the correction coefficient CF and output to the correction unit 88.

[0098] As an example, such as Figure 13 As shown, the correction unit 88 sets the target Raman spectrum 37T as the corrected target Raman spectrum 37TC by multiplying the correction coefficient CF by all intensity values ​​of the target Raman spectrum 37T.

[0099] As an example, such as Figure 14 As shown, the prediction unit 89 inputs the Raman spectrum data 37TC of the target object into the antibody concentration prediction model 82, and outputs the prediction result 95 from the antibody concentration prediction model 82. The prediction result 95 includes the predicted value of the concentration of antibody 16 in the target suspension 18T.

[0100] As an example, such as Figure 15 As shown, the antibody concentration prediction model 82 is constructed by a neural network 125. Therefore, the antibody concentration prediction model 82 is also an example of the "machine learning model" involved in the technology of this invention. As is well known, the neural network 125 has an input layer 126, an intermediate layer (also called a hidden layer) 127, and an output layer 128. These input layers 126, intermediate layers 127, and output layers 128 each have multiple nodes ND. Coefficients representing the binding strength of each node ND are set between the nodes ND of the input layer 126 and the nodes ND of the intermediate layer 127, between the nodes ND within the intermediate layer 127, and between the nodes ND of the intermediate layer 127 and the nodes ND of the output layer 128. Appropriate activation functions such as linear functions and ReLU (Rectified Linear Unit) functions are set in the nodes ND of the output layer 128.

[0101] The intensity values ​​of each wavenumber of the Raman spectrum data 37TC of the calibration target are input to each node ND of the input layer 126. Furthermore, the predicted concentration of antibody 16 is output from the node ND of the output layer 128. Additionally, the antibody concentration prediction model 82 is not limited to the illustrated neural network 125, but can be other machine learning models such as decision trees, gradient boosting decision trees, random forests, support vector machines, and Naive Bayes.

[0102] As an example, such as Figure 16 As shown, during the learning phase of antibody concentration prediction model 82, teaching data (also referred to as learning data or training data) 130 is used. Teaching data 130 is a set of reference Raman spectral data 37STL and the forward antibody concentration 95CA. Reference Raman spectral data 37STL is obtained by... Figure 8The data were obtained by measuring the Raman scattered light RSL emitted from a reference suspension 18ST containing at least antibody 16 under the reference measurement conditions shown. The positive antibody concentration 95CA is the concentration of antibody 16 in the reference suspension 18ST, which will be the source of the reference Raman spectrum data 37STL for study, after it has been introduced into a high-performance liquid chromatography (HPLC) apparatus and measured by the mass analysis function of the HPLC apparatus.

[0103] During the learning phase, the reference Raman spectral data 37STL is input into the antibody concentration prediction model 82, and the learning prediction result 95L is output from the antibody concentration prediction model 82. Next, based on the comparison between the learning prediction result 95L and the correct antibody concentration 95CA, a loss calculation is performed on the antibody concentration prediction model 82 using a loss function. Then, based on the result of the loss calculation, the coefficients (coefficients of node ND) of the antibody concentration prediction model 82 are updated, and the antibody concentration prediction model 82 is updated according to the updated settings.

[0104] During the learning phase, while changing the teaching data 130, the above-described series of processes are repeatedly performed: inputting the learning reference Raman spectrum data 37STL into the antibody concentration prediction model 82, outputting the learning prediction result 95L from the antibody concentration prediction model 82, performing loss calculations, updating settings, and updating the antibody concentration prediction model 82. The repetition of the above-described series of processes ends when the prediction accuracy of the learning prediction result 95L relative to the correct antibody concentration 95CA reaches a preset level. Thus, the antibody concentration prediction model 82 with the preset prediction accuracy is stored in the storage device 70 and used in the prediction unit 89. This repetition of the above-described series of processes is an example of "correction" according to the technology of this invention. Alternatively, the learning can end after repeating the above-described series of processes a preset number of times, regardless of the prediction accuracy of the learning concentration prediction result 95L relative to the correct antibody concentration 95CA.

[0105] The display control unit 90 receives the prediction result 95 from the prediction unit 89 and performs the following control: (As an example) Figure 17The prediction result display screen 135 is displayed on the monitor 41. The prediction result display screen 135 shows the predicted value of antibody 16 concentration from prediction result 95. Furthermore, a save button 136 and an confirm button 137 are provided at the bottom of the prediction result display screen 135. When the save button 136 is selected, the predicted value is stored in the storage device 70 in association with the object Raman spectral data 37T. When the confirm button 137 is selected, the display of the prediction result display screen 135 is cleared. Alternatively, for example, the antibody 16 concentration can be predicted based on the antibody concentration prediction model 82 at multiple time points at constant intervals, obtaining predicted values ​​of antibody 16 concentration at multiple time points. These predicted values ​​are then plotted on a graph with time as the horizontal axis, and the time series change of the predicted antibody 16 concentration is displayed on the prediction result display screen 135.

[0106] Next, as an example, refer to Figure 18 The flowchart shown illustrates the function of the structure described above. Figure 6 As shown, the CPU 72 of the information processing device 40 functions as the acquisition unit 85, the RW control unit 86, the instruction receiving unit 87, the correction unit 88, the prediction unit 89, and the display control unit 90 by starting the working program 80.

[0107] First, in the acquisition unit 85, object Raman spectrum data 37T from Raman spectrometer 12 is acquired (step ST100). The object Raman spectrum data 37T is output from the acquisition unit 85 to the RW control unit 86, and stored in the storage device 70 under the control of the RW control unit 86 (step ST110).

[0108] Under the control of the display control unit 90 Figure 11 The flow cell selection screen 115 shown is displayed on the monitor 41 (step ST120). User U selects the flow cell 11 used when the object Raman spectrum data 37T was obtained using the drop-down menu 116, and then selects the OK button 117. Thus, as shown... Figure 12 As shown, the designated information 120 is received by the instruction receiving unit 87 (step ST130). The designated information 120 is output from the instruction receiving unit 87 to the RW control unit 86.

[0109] Under the control of the RW control unit 86, the target Raman spectrum data 37T is read from the storage device 70. Furthermore, the correction coefficient CF corresponding to the flow cell 11 specified by the designation information 120 is obtained by reading the correction coefficient information 81 from the storage device 70 (step ST140). The target Raman spectrum data 37T and the correction coefficient CF are output from the RW control unit 86 to the correction unit 88.

[0110] In the calibration section 88, such asFigure 13 As shown, the Raman spectrum data 37TC of the target object is generated based on the correction coefficient CF and the Raman spectrum data 37T of the target object (step ST150). The Raman spectrum data 37TC of the target object is output from the correction unit 88 to the prediction unit 89.

[0111] In the prediction section 89, such as Figure 14 As shown, the Raman spectrum data 37TC of the calibration target is input into the antibody concentration prediction model 82, thereby outputting the prediction result 95 from the antibody concentration prediction model 82 (step ST160). The prediction result 95 is output from the prediction unit 89 to the display control unit 90.

[0112] like Figure 17 As shown, under the control of the display control unit 90, the prediction result display screen 135 is displayed on the display 41 (step ST170), and the user U can view the predicted value of the concentration of antibody 16 in the prediction result 95.

[0113] User U makes various decisions based on the predicted values ​​displayed on the prediction results display screen 135. For example, consider the case where conditions such as cell culture conditions are set for antibody production based on small-scale equipment. In this case, if the predicted value is worse than the target value, user U decides to stop the current experiment and switch to an experiment based on the new conditions. Furthermore, consider the case where conditions are set and mass production is carried out based on large-scale equipment. In this case, if the predicted value is worse than the target value, user U decides to interrupt mass production and perform maintenance on the culture tanks in culture unit 13 or the various chromatographic devices in purification unit 14.

[0114] As described above, the information processing device 40 includes an acquisition unit 85, an RW control unit 86, a correction unit 88, and a prediction unit 89. The prediction unit 89 uses an antibody concentration prediction model 82 (learned from the reference Raman spectrum data 37STL) corrected based on the reference Raman spectrum data 37ST obtained from the reference suspension 18ST under a reference measurement environment. The acquisition unit 85 acquires object Raman spectrum data 37T from an object suspension 18T with an unknown antibody concentration under an object measurement environment different from the reference measurement environment. The RW control unit 86 acquires a correction coefficient CF derived from a comparison of the reference Raman spectrum data 37ST and the reference Raman spectrum data 37R corresponding to the object Raman spectrum data 37T, obtained from the reference suspension 18ST under the object measurement environment. The correction unit 88 corrects the object Raman spectrum data 37T based on the correction coefficient CF to generate corrected object Raman spectrum data 37TC. The prediction unit 89 applies the Raman spectral data 37TC of the calibration object to the antibody concentration prediction model 82, enabling the antibody concentration prediction model 82 to predict the concentration of antibody 16.

[0115] Differences in Raman spectral data 37 originating from the measurement environment can be eliminated by correction based on the correction factor CF. Therefore, even when the measurement environment of Raman scattered light RSL changes, highly reliable predictions of the concentration 95 of antibody 16 in the target suspension 18T can be obtained.

[0116] like Figure 10 As shown, the correction factor CF is derived from the ratio of the intensity values ​​of the reference Raman spectrum data 37ST to the reference Raman spectrum data 37R corresponding to the components in the reference suspension 18ST. Therefore, a reasonable correction factor CF can be easily derived. Alternatively, a machine learning model that outputs the correction factor CF based on the inputs of the reference Raman spectrum data 37ST and the reference Raman spectrum data 37R can also be used.

[0117] And, as Figure 10 As shown, the correction factor CF is derived after removing the ratio of the intensity value corresponding to the intensity value of the measuring instrument for Raman scattered light RSL, which is here the optical system 58 of the flow cell 11.

[0118] As described above, when using flow cell 11, in the Raman spectral data 37, not only were peaks corresponding to the components in the target suspension 18T observed, but also peaks originating from the optical system 58 of flow cell 11 were observed. The peaks originating from this optical system 58 and the peaks corresponding to the components in the target suspension 18T appear in different ways. Therefore, as... Figure 10 As shown in Figure 111, the ratio of the intensity value of the reference Raman spectrum data 37ST corresponding to the optical system 58 to the intensity value of the reference Raman spectrum data 37R corresponding to the optical system 58 deviates from the ratio of the intensity value of the reference Raman spectrum data 37ST corresponding to the components in the reference suspension 18ST to the intensity value of the reference Raman spectrum data 37R corresponding to the components in the reference suspension 18ST. Therefore, if the correction factor CF is derived by removing the ratio of the intensity values ​​corresponding to the optical system 58 of this flow cell 11, a more appropriate correction factor CF can be derived.

[0119] like Figure 8 As shown, the reference measurement environment and the object measurement environment differ in the presence or absence of flow cell 11 in the measurement instrument for Raman scattering light RSL. Therefore, a more reliable correction factor CF can be derived.

[0120] like Figure 7 As shown, there are multiple object measurement environments, each differing in the type of flow cell 11 used. Furthermore, the correction coefficient CF is pre-stored in the storage device 70 for each of the multiple object measurement environments. Moreover, as... Figure 11 and Figure 12As shown, the receiving unit 87 receives a specification of the type of flow cell 11 used. The RW control unit 86 obtains the correction coefficient CF corresponding to the specification by reading from the storage device 70.

[0121] Therefore, it can handle multiple test environments without any problems. Furthermore, it eliminates the labor and time required to prepare multiple test environments and, consequently, antibody concentration prediction models 82 for each type of flow cell 11. Moreover, the teaching data 130 required for the antibody concentration prediction model 82 consists only of the learning reference Raman spectral data 37STL and the forward antibody concentration 95CA obtained in the reference test environment. Therefore, the teaching data 130 can be prepared simply.

[0122] The Raman scattered light RSL of the optical system 58 in the flow cell 11 has at least one different component: the distance D between the incident surface of the transparent plate 60 (the second surface 62) and the wall surface 63A of the flow path 52 opposite the second surface 62; the reflectivity R of the wall surface 63A; and the focal length FL of the optical system 58. This allows for seamless handling of variations in the object measurement environment caused by these differences in distance D, reflectivity R, and focal length FL.

[0123] Concentration is the most commonly used indicator when the physicochemical characteristics of antibody 16 are known. Therefore, as in this example, predicting the concentration of antibody 16 based on its state would allow user U to easily understand the physicochemical characteristics of antibody 16.

[0124] Biopharmaceuticals containing antibody 16 are called antibody pharmaceuticals, and they are widely used not only in the treatment of chronic diseases such as cancer, diabetes, and rheumatoid arthritis, but also in the treatment of rare diseases such as hemophilia and Crohn's disease. Therefore, this example of using a protein as antibody 16 can promote the development of antibody pharmaceuticals that are widely used in the treatment of various diseases.

[0125] Raman scattering light (RSL) readily reflects information about the functional groups of amino acids derived from proteins. Therefore, as in this example, by setting the electromagnetic wave as Raman scattering light (RSL), Raman spectral data 37 that accurately reflects the concentration and other physical properties of antibody 16 as a protein can be obtained.

[0126] like Figure 15 and Figure 16 As shown, antibody concentration prediction model 82 is a machine learning model that learns using the benchmark Raman spectral data 37STL as teaching data. Machine learning models are generally used for predicting unknown parameters and can improve prediction accuracy to a certain level through learning. Therefore, it is easy to generate antibody concentration prediction model 82 with high prediction accuracy.

[0127] (Modified example) In the correction coefficient information 81 of the first embodiment described above, an example is shown where a correction coefficient CF is registered for a flow cell 11, but it is not limited to this. As an example, such as... Figure 19 The correction factor information 140 shown can also be used to register correction factors CF for each wavenumber range of the Raman spectral data 37 for a flow cell 11. And, as an example, such as... Figure 20 The correction factor information 142 shown can also be used to register correction factors CF for each wavenumber of a flow cell 11. In this way, by setting the correction factors CF in detail according to each wavenumber range and each wavenumber, a more meticulous correction can be performed on the object Raman spectral data 37T.

[0128] [Second Implementation] In the first embodiment described above, such as Figure 16 The example shown is teaching data 130 containing reference Raman spectral data 37STL for learning, but it is not limited to this. As an example, such as... Figure 21 As shown, in addition to the teaching data 130 containing the reference Raman spectrum data 37STL for learning, the teaching data 130 containing the calibration target Raman spectrum data 37TCL for learning can also be used. The calibration target Raman spectrum data 37TCL for learning is generated by correcting the previously measured target Raman spectrum data 37T according to the correction factor CF.

[0129] Thus, in the second embodiment, the antibody concentration prediction model 82 is learned using the Raman spectral data 37TCL of the calibration target. Therefore, a large amount of teaching data 130 can be ensured. As a result, the prediction accuracy of the antibody concentration prediction model 82 can be improved.

[0130] [Example] First of all, Figure 8 In a reference measurement environment where the sensor section 35 of the Raman spectrometer 12 was directly immersed in the reference suspension 18ST within container 100A, the Raman scattered light RSL from the reference suspension 18ST was measured. Thus, reference Raman spectral data 37ST was obtained. Figure 16 As shown, using the reference Raman spectral data 37ST obtained in this way as the teaching data 130 for learning reference Raman spectral data 37STL, the antibody concentration prediction model 82 is trained, and an antibody concentration prediction model 82 with prediction accuracy reaching the set level is obtained.

[0131] like Figure 1 As shown, a flow cell 11 is provided on the outflow path 26 of the concentration section 15, and the sensor section 35 of the Raman spectrometer 12 is connected to the flow cell 11 to measure the Raman scattered light RSL of the target suspension 18T, thereby obtaining the target Raman spectrum data 37T. Figure 13As shown, it is applicable through Figure 10 The correction coefficients exported by the shown correction coefficient export process 101 are applicable to the Raman spectral data 37T of the object, generating the corrected Raman spectral data 37TC of the object.

[0132] Then, as Figure 14 As shown, the Raman spectrum data 37TC of the target object was input into the antibody concentration prediction model 82, and the prediction result 95 was output from the antibody concentration prediction model 82. On the other hand, the concentration of antibody 16 in the target suspension 18T was measured using the quality analysis function of the HPLC device. The concentration of antibody 16 was predicted based on the antibody concentration prediction model 82 and measured based on the HPLC device at various time points, including the start of concentration in the concentration section 15, intermediate 1, intermediate 2, final, and recovery (referred to as recovery) of the buffer solution used for washing.

[0133] The predicted values ​​of antibody 16 concentration based on antibody concentration prediction model 82 at each time point in the embodiment are compared with the measured values ​​of antibody 16 concentration based on the HPLC device (marked as offline analysis values ​​in the figure). Figures 24-26 The comparison chart 150 (also the same) is shown in Figure 22 The average error (hereinafter referred to as the average error) between the predicted value and the measured value at each time point is 3.7%.

[0134] [Comparative Example] As an example, such as Figure 23 Table 152 shows Comparative Examples 1, 2, and 3. Comparative Example 1 involves directly inputting the target Raman spectrum data 37T into the antibody concentration prediction model 82 without performing correction based on the correction coefficient CF. Comparative Example 2 involves performing the derivative transformation (including 15-point data smoothing) described in Japanese Patent Publication No. 2022-552876 on the target Raman spectrum data 37T. Comparative Example 3 involves performing the derivative transformation (including 15-point data smoothing) and standard normalization on the target Raman spectrum data 37T as described in Japanese Patent Publication No. 2022-552876.

[0135] The comparison chart 160 shows the predicted values ​​of antibody 16 concentration at each time point based on antibody concentration prediction model 82 and the measured values ​​of antibody 16 concentration based on the HPLC device in Comparative Example 1. Figure 24 In the middle, at any given time point, the predicted value was lower than the measured value. This is consistent with... Figure 10 The results show that the correction factor CF is 1 or higher. The average error in this case is 6.0%, which is worse than the previous example.

[0136] Chart 161 compares the predicted values ​​of antibody 16 concentration at each time point based on antibody concentration prediction model 82 with the measured values ​​of antibody 16 concentration based on the HPLC device in Comparative Example 2. Figure 25 In this case, the average error was 16.9%, a significant deterioration compared to the previous example.

[0137] Chart 162 compares the predicted values ​​of antibody 16 concentration at each time point based on antibody concentration prediction model 82 with the measured values ​​of antibody 16 concentration based on the HPLC device in Comparative Example 3. Figure 26 In this case, the average error was 12.9%, which, like Comparative Example 1, was a significant deterioration compared to the Example 1.

[0138] As can be seen from the above, the embodiments using the technology of the present invention have more effective effects than comparative examples 1 to 3.

[0139] As a state, the concentration of antibody 16 is predicted, but it is not limited to this. The concentration of antibody 16 aggregates 17 can also be predicted. Furthermore, instead of concentration, or in addition to this, purity, density, etc., can also be predicted. Purity is calculated, for example, with the sum of the amount of antibody 16 and the amount of dopant as the denominator and the amount of antibody 16 as the numerator. For example, more than two states, such as concentration and density, can be predicted. Moreover, it is not limited to quantitative indicators such as concentration and purity, but can also be qualitative indicators such as the quality level of antibody 16 (which can be two stages of good and bad or five stages from 1 to 5, etc.).

[0140] As a state, it can also be the concentration of living cells, glucose concentration, glutamate concentration, amino acid concentration, solvent component concentration, additive concentration, lactic acid concentration, ammonia concentration, concentration of cell-derived protein / cell-derived DNA, concentration of other cell metabolites, concentration of antibody fragments, concentration of charge isomers, etc.

[0141] The substances for which Raman spectroscopy data 37 is measured are not limited to antibodies 16, etc. They can also be proteins, peptides, nucleic acids (DNA, RNA (ribonucleic acid)), lipids, viruses, viral subunits, and virus-like particles, in addition to antibodies 16.

[0142] The components are not limited to antibody 16. They can also be cytokines (interferon, interleukin, etc.) or hormones (insulin, glucagon, follicle-stimulating hormone, erythropoietin, etc.), growth factors (IGF (Insulin-Like Growth Factor)-1, bFGF (Basic Fibroblast Growth Factor), etc.), coagulation factors (Factors 7, 8, 9, etc.), enzymes (lysosomal enzymes, DNA (Deoxyribonucleic Acid) degrading enzymes, etc.), Fc (Fragment Crystallizable) fusion proteins, receptors, albumin, and protein vaccines. Furthermore, antibody 16 also includes bispecific antibodies, antibody-drug conjugates, low-molecular-weight antibodies, and glycan-modified antibodies.

[0143] Electromagnetic waves are not limited to Raman scattered light (RSL), therefore spectral data are not limited to Raman spectral data.37 They can also be infrared absorption spectral data, near-infrared absorption spectral data, nuclear magnetic resonance spectral data, ultraviolet-visible (UV-Vis) spectral data, or fluorescence spectral data.

[0144] The antibody concentration prediction model 82 is not limited to a machine learning model. It can also be a model generated through multivariate analysis and statistical analysis. Examples of multivariate analysis and statistical analysis include linear regression, multiple regression, principal component regression, partial least squares regression, logistic regression, Lasso regression, ridge regression, support vector regression, and Gaussian process regression. In models generated through such multivariate analysis and statistical analysis, determining the regression coefficients based on at least two teaching data 130 is an example of the "correction" involved in the technology of this invention.

[0145] In addition, the different measurement environments that replace the flow cell 11 may vary, including the presence or type of the optical components, the shape of the flow path 52, temperature, humidity, and the flow rate and velocity of the target suspension 18T. Specifically, the optical components may be optical fibers for guiding the excitation light EL and Raman scattering light RSL, or the light source of the excitation light EL, and the optical system 58, etc., which are disposed in the sensor section 35 of the Raman spectrometer 12. Specifically, the shape of the flow path 52 may be a circular cross-section, an elliptical cross-section, or an overall shape such as a straight line, a U-shape, or an L-shape.

[0146] The flow path 52 of the flow cell 11 can be U-shaped. Furthermore, the shape of the flow cell 11 is not limited to a cylindrical shape; it can also be a square cylinder. The cross-sectional shape of the flow path 52 is not limited to a circle; it can be elliptical or rectangular. Moreover, the flow cell 11 can also be formed from composite materials such as carbon fiber reinforced resin.

[0147] The fluid is not limited to the purification solution 18 and the mixture 18M of the purification solution 18 and the concentrate. It can be the cell culture medium before the cells are removed by the decellularization filter in the culture section 13, or the culture supernatant after the cells are removed. It can also be the cell culture medium (culture medium) that does not contain biomolecules supplied to the culture section 13. It can also be the purification solution introduced into or sent from various chromatographic devices. It can also be the purification solution after the treatment to inactivate the virus has ended.

[0148] Fluids are not limited to those involved in biopharmaceuticals; for example, they can also include river water collected for water pollution investigations. Furthermore, fluids are not limited to liquids; they can also be gases.

[0149] In the above embodiments, an example is illustrated where the sensor section 35 of the Raman spectrometer 12, which is the measurement sensor, is provided within the manufacturing process, and the measurement is performed within the manufacturing process. However, this is not a limitation. The technology of the present invention can also be applied to offline sensing, which uses a measurement sensor separate from the manufacturing process and performs the measurement outside the manufacturing process.

[0150] The optical element constituting the optical system 58 is not limited to the spherical lens 59 illustrated. It can also be a hemispherical lens, a plano-convex lens, a biconvex lens, or a cylindrical lens. Furthermore, the optical element is not limited to the disk-shaped transparent plate 60 with a first surface 61 and a second surface 62 that are parallel to each other, as illustrated. It can also be a transparent plate with a first surface and a planar second surface that mimic the shape of the exit surface of a spherical lens, a plano-convex lens, or a biconvex lens. In addition, the optical system 58 may not have a transparent plate, or it may be constituted solely by the spherical lens 59, a hemispherical lens, or the like.

[0151] like Figure 1 As shown, the information processing device 40 can be a personal computer installed in a pharmaceutical facility or a server computer installed in a data center independent of the pharmaceutical facility.

[0152] In the case where the information processing device 40 is composed of a server computer, the target Raman spectroscopy data 37T is sent from personal computers installed in various pharmaceutical facilities to the server computer via a network such as the Internet. The server computer transmits various screens, such as the flow pool selection screen 115 and the prediction result display screen 135, to the personal computer in a format for web page transmission screen data created using a markup language such as XML (Extensible Markup Language). The personal computer reproduces the screens displayed on the web browser based on the screen data and displays them on the monitor. Alternatively, other data description languages ​​such as JSON (Javascript Object Notation) can be used instead of XML.

[0153] The hardware structure of the computer constituting the information processing apparatus 40 according to the technology of the present invention can be modified in various ways. For example, to improve processing power and reliability, the information processing apparatus 40 can be composed of multiple computers that are separate as hardware. For example, two computers can be used to distribute the functions of the correction unit 88 and the prediction unit 89. In this case, the information processing apparatus 40 is composed of two computers.

[0154] Thus, the hardware structure of the computer in the information processing device 40 can be appropriately modified according to the performance requirements such as processing power, security, and reliability. Furthermore, not limited to hardware, applications such as the operating program 80 can, for the purpose of ensuring security and reliability, be dual-stored or distributed across multiple storage devices.

[0155] In the above embodiments, for example, as the hardware structure of the processing unit that performs various processes, such as the acquisition unit 85, the RW control unit 86, the instruction receiving unit 87, the correction unit 88, the prediction unit 89, and the display control unit 90, various processors as shown below can be used. As described above, the various processors include, in addition to the general-purpose processor CPU 72 that executes software (working program 80) to function as various processing units, processors such as FPGA (Field Programmable Gate Array) whose circuit structure can be changed after manufacturing, i.e., Programmable Logic Device (PLD), and processors such as ASIC (Application Specific Integrated Circuit) with circuit structures specifically designed for performing specific processes, i.e., dedicated circuits.

[0156] A processing unit can consist of one of these various processors, or it can consist of a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs and / or a combination of a CPU and an FPGA). Furthermore, multiple processing units can also be composed of a single processor.

[0157] As examples of a single processor comprising multiple processing units, firstly, there are methods such as client and server computers, where a single processor is constructed by combining one or more CPUs with software, and this processor functions as multiple processing units. Secondly, there are methods such as System-on-Chip (SoC), where a single IC (Integrated Circuit) chip is used to implement the functions of the entire system, which includes multiple processing units. In this way, various processing units are constructed as hardware structures using one or more of the aforementioned processors.

[0158] Furthermore, the hardware structure of these various processors, more specifically, can be a circuit composed of circuit elements such as semiconductor components.

[0159] Based on the above records, one can master the techniques described in the following notes.

[0160] [Note 1] An information processing device that predicts the state of a target component in a suspension based on spectral data obtained by measuring electromagnetic waves emitted from a suspension in which biomolecules are dispersed as components, the information processing device comprising a processor. The processor performs the following processing: A state prediction model corrected based on reference spectral data obtained from a reference suspension was used in a reference measurement environment. In an object measurement environment different from the aforementioned benchmark measurement environment, object spectral data are obtained from an object suspension in which the state of the object components is unknown. A correction coefficient is obtained by comparing the reference spectral data and the reference spectral data corresponding to the spectral data of the object with the reference spectral data obtained from the reference suspension under the object measurement environment; Corrected object spectral data is generated by correcting the object spectral data according to the correction coefficient; and By applying the spectral data of the corrected object to the state prediction model, the state prediction model can predict the state of the object components.

[0161] [Note 2] In the information processing apparatus according to Note 1, wherein, The correction factor is derived based on the ratio of the intensity values ​​of the reference spectral data corresponding to the component in the reference suspension to the intensity values ​​of the reference spectral data.

[0162] [Appendix 3] In the information processing apparatus according to Appendix 1 or 2, wherein, The correction factor is derived after removing the ratio of the intensity value corresponding to the measuring instrument of the electromagnetic wave.

[0163] [Note 4] In the information processing apparatus according to Note 3, wherein, The measuring instrument is a flow cell having a flow path for the suspension to flow and an optical system for taking in the electromagnetic waves. The intensity value corresponding to the measuring instrument is the intensity value corresponding to the optical system.

[0164] [Note 5] The information processing apparatus according to any one of Notes 1 to 4, wherein, The reference measurement environment and the object measurement environment differ in either the presence or use of the electromagnetic wave measuring instrument or the type thereof.

[0165] [Note 6] In the information processing apparatus according to Note 5, wherein, The measuring instrument is a flow cell having a flow path for the suspension to flow and an optical system for taking in the electromagnetic waves.

[0166] [Note 7] In the information processing apparatus according to Note 6, wherein, There are multiple measurement environments for the objects mentioned above. The measurement environments for the various objects differ in the type of flow cell used.

[0167] [Note 8] In the information processing apparatus according to Note 7, wherein, The correction coefficients are pre-stored in the storage unit for each of the multiple environments in which the objects are measured. The processor performs the following processing: The designation of the type of flow cell used is received; and The correction coefficient corresponding to the specified value is obtained by reading from the storage unit.

[0168] [Note 9] In the information processing apparatus according to Note 7 or 8, wherein, The distance between the incident surface of the electromagnetic wave of the optical system of the flow cell and the wall of the flow path opposite the incident surface, the reflectivity of the wall, and the focal length of the optical system are different from at least one of the following:

[0169] [Note 10] The information processing apparatus according to any one of Notes 1 to 9, wherein, The state prediction model is corrected based on the spectral data of the object being corrected.

[0170] [Note 11] The information processing apparatus according to any one of Notes 1 to 10, wherein, The state of the object component is the concentration of the object component.

[0171] [Note 12] The information processing apparatus according to any one of Notes 1 to 11, wherein, The object is composed of protein.

[0172] [Note 13] In the information processing apparatus according to Note 12, wherein, The protein in question is an antibody.

[0173] [Note 14] The information processing apparatus according to any one of Notes 1 to 13, wherein, The electromagnetic wave is Raman scattered light.

[0174] [Note 15] The information processing apparatus according to any one of Notes 1 to 14, wherein, The state prediction model is a machine learning model that learns using the benchmark spectral data as teaching data.

[0175] The technology of the present invention can also be appropriately combined with the various embodiments and / or variations described above. Furthermore, it is not limited to the embodiments described above; of course, various structures can be adopted as long as the spirit is not departed from. Moreover, the technology of the present invention relates not only to programs, but also to storage media that do not temporarily store programs and computer program articles containing programs.

[0176] The descriptions and illustrations shown above are detailed explanations of the parts involved in the technology disclosed herein, and are merely one example of the technology disclosed herein. For example, the descriptions related to the structure, function, role, and effect described above are only one example of the structure, function, role, and effect of the parts involved in the technology disclosed herein. Therefore, it is self-evident that unnecessary parts may be deleted, new elements may be added, or substitutions may be made to the descriptions and illustrations shown above without departing from the spirit of the technology disclosed herein. Furthermore, in order to avoid complexity and facilitate understanding of the parts involved in the technology disclosed herein, descriptions related to technical common sense that do not require special explanation have been omitted from the descriptions and illustrations shown above, based on the premise that the technology disclosed herein can be implemented.

[0177] In this specification, "A and / or B" has the same meaning as "at least one of A and B". That is, "A and / or B" means that it can be only A, only B, or a combination of A and B. Furthermore, in this specification, when three or more things are connected and expressed using "and / or", the same way of thinking applies as with "A and / or B".

[0178] All documents, patent applications and technical standards described in this specification are incorporated herein by reference to the same extent as those specifically and individually described herein.

Claims

1. An information processing apparatus that predicts the state of a target component in a suspension based on spectral data obtained by measuring electromagnetic waves emitted from a suspension in which biomolecules are dispersed as components, the information processing apparatus comprising a processor. The processor performs the following processing: A state prediction model corrected based on reference spectral data obtained from a reference suspension under reference measurement conditions was used. In an object measurement environment different from the aforementioned benchmark measurement environment, object spectral data are obtained from an object suspension in which the state of the object components is unknown. A correction coefficient is obtained by comparing the reference spectral data with a reference spectral data corresponding to the spectral data of the object, wherein the reference spectral data is obtained from the reference suspension under the object measurement environment; The corrected object spectral data is generated by correcting the object spectral data using the correction coefficients; and By applying the spectral data of the corrected object to the state prediction model, the state prediction model can predict the state of the object components.

2. The information processing apparatus according to claim 1, wherein, The correction factor is derived based on the ratio of the intensity values ​​of the reference spectral data and the reference spectral data corresponding to the component in the reference suspension.

3. The information processing apparatus according to claim 1, wherein, The correction factor is derived after removing the ratio of the intensity value corresponding to the measuring instrument of the electromagnetic wave.

4. The information processing apparatus according to claim 3, wherein, The measuring instrument is a flow cell having a flow path for the suspension to flow and an optical system for taking in the electromagnetic waves. The intensity value corresponding to the measuring instrument is the intensity value corresponding to the optical system.

5. The information processing apparatus according to claim 1, wherein, The reference measurement environment and the object measurement environment differ in either the presence or use of the electromagnetic wave measuring instrument or the type thereof.

6. The information processing apparatus according to claim 5, wherein, The measuring instrument is a flow cell having a flow path for the suspension to flow and an optical system for taking in the electromagnetic waves.

7. The information processing apparatus according to claim 6, wherein, There are multiple measurement environments for the objects mentioned above. The measurement environments for the various objects differ in the type of flow cell used.

8. The information processing apparatus according to claim 7, wherein, The correction coefficients are pre-stored in the storage unit for each of the multiple environments in which the objects are measured. The processor performs the following processing: The designation of the type of flow cell used is received; and The correction coefficient corresponding to the specified value is obtained by reading from the storage unit.

9. The information processing apparatus according to claim 7, wherein, The distance between the incident surface of the electromagnetic wave and the wall of the flow path opposite to the incident surface of the optical system of the flow cell, the reflectivity of the wall, and the focal length of the optical system are different at least one of the following:

10. The information processing apparatus according to claim 1, wherein, The state prediction model is corrected using the spectral data of the object being corrected.

11. The information processing apparatus according to claim 1, wherein, The state of the object component is the concentration of the object component.

12. The information processing apparatus according to claim 1, wherein, The object is composed of protein.

13. The information processing apparatus according to claim 12, wherein, The protein in question is an antibody.

14. The information processing apparatus according to claim 1, wherein, The electromagnetic wave is Raman scattered light.

15. The information processing apparatus according to claim 1, wherein, The state prediction model is a machine learning model that learns using the benchmark spectral data as teaching data.

16. A method of operating an information processing apparatus, comprising predicting the state of a target component in a suspension based on spectral data obtained by measuring electromagnetic waves emitted from a suspension in which biomolecules are dispersed as components, the method comprising the following steps: A state prediction model corrected based on reference spectral data obtained from a reference suspension under reference measurement conditions was used. In an object measurement environment different from the aforementioned benchmark measurement environment, object spectral data are obtained from an object suspension in which the state of the object components is unknown. A correction coefficient is obtained by comparing the reference spectral data with a reference spectral data corresponding to the spectral data of the object, wherein the reference spectral data is obtained from the reference suspension under the object measurement environment; Corrected object spectral data is generated by correcting the object spectral data according to the correction coefficient; and By applying the spectral data of the corrected object to the state prediction model, the state prediction model can predict the state of the object components.

17. An operating procedure for an information processing device, the operating procedure of which predicts the state of a target component in a suspension based on spectral data obtained by measuring electromagnetic waves emitted from a suspension in which biomolecules are dispersed as components, and causes a computer to perform processing including the following steps: A state prediction model corrected based on reference spectral data obtained from a reference suspension under reference measurement conditions was used. In an object measurement environment different from the aforementioned benchmark measurement environment, object spectral data are obtained from an object suspension in which the state of the object components is unknown. A correction coefficient is obtained by comparing the reference spectral data with a reference spectral data corresponding to the spectral data of the object, wherein the reference spectral data is obtained from the reference suspension under the object measurement environment; Corrected object spectral data is generated by correcting the object spectral data according to the correction coefficient; and By applying the spectral data of the corrected object to the state prediction model, the state prediction model can predict the state of the object components.

Images