Morphology analysis device, morphology analysis method and morphology analysis program

Abstract

Provided are a morphology analysis device, a morphology analysis method, and a morphology analysis program with which it is possible to recognize the morphology of the constituent elements of molecules. A morphology analysis device 200 for analyzing the morphology of molecules in a solution comprises: a reference data storage unit 234 that stores reference data, which is molecular shape data for a pre-identified test molecule; a target data storage unit 235 that stores target data, which is molecular shape data for the test molecule serving as an analysis target; and a morphology estimation unit 237 that estimates the morphology of the constituent elements of a molecule in the target data by identifying the state of physical quantities constituting the molecular shape of the target data with respect to the reference data. Both the reference data and the target data have a molecular level resolution.

Description

Morphological analysis device, morphological analysis method, and morphological analysis program

[0001] The present invention relates to a morphology analysis device, a morphology analysis method, and a morphology analysis program for analyzing the morphology of molecules in a solution.

[0002] Biopharmaceuticals are proteins, containing large molecules with complex structures as their primary components. Unlike small molecule drugs, biopharmaceuticals cannot be produced as chemically uniform proteins. Therefore, their efficacy and safety are significantly affected by factors such as conformation, heterogeneity of glycan structures, and nonspecific associations. Therefore, to ensure reliability, post-translational modification, aggregation analysis, and analysis of three-dimensional structural changes are required. The production of biopharmaceuticals requires large-scale capital investment, and quality control is extremely difficult (Non-Patent Document 1).

[0003] When biopharmaceuticals are administered, not only are the dosages large, but thousands of times the amount of active ingredient is required for a single dose. In addition, the structure and quality of the final product must be evaluated, and the comparability of the products, including various parameters, must also be ensured.

[0004] Even if biopharmaceuticals are produced using the same gene sequence, if the conformation, glycan structure, or association is not uniform, the drug is not the same. Therefore, to ensure quality within certain specifications, conditions such as cultivation, purification, and concentration must be strictly controlled.

[0005] In particular, when producing antibodies using CHO cell lines, long-term culture reduces antibody production capacity and leads to the excretion of impurities. This requires constantly culturing fresh cells. Furthermore, expression regulation is difficult to control, which can result in the production of low-quality antibodies. Therefore, comprehensive quality control requires not only conventional analysis of chemical properties and comprehensive physicochemical properties, but also confirmation of structural changes and fluctuations.

[0006] However, existing biopharmaceutical analysis and characterization processes require a long time, and furthermore, analytical methods must be developed in advance according to the state of the sample. Furthermore, three-dimensional structural analysis techniques, which are extremely limited, have the crucial limitation that they cannot be measured in solution. Therefore, it is not possible to measure or analyze the properties related to three-dimensional structure of the drug substance under the desired conditions or state. In other words, it is not possible to know not only the conformation of the drug substance molecule at that time, but also the distribution of its mobility. As such, molecular morphological analysis that is effective for process control of pharmaceutical development or manufacturing has not been performed.

[0007] Conventionally, there are methods for observing biopolymers by freezing samples, such as cryo-electron microscopy, but the mobility of biopolymers is not apparent because the sample is frozen. Another known method uses a database to obtain a three-dimensional distribution of electron density based on the coordinates of atoms that make up proteins with known structures, but it is not possible to recognize the mobile structure from the obtained three-dimensional distribution (Patent Document 1). Furthermore, it is obvious that neither of these methods can provide observation results under given conditions, which are essential in the process of pharmaceutical development or quality control.

[0008] "Current Status and Issues of the Biopharmaceutical Industry", Junichi Matsuzaki, Journal of Biotechnology, 91(9), 495-498, 2013

[0009] Japanese Patent Application Laid-Open No. 2005-250721

[0010] As described above, there is a need for technology to observe three-dimensional structure-related information that can be applied to quality control, including the formulation development and manufacturing processes of biopharmaceuticals. However, conventional technology is unable to identify the electron density of components that reflect conformational changes and mobility of biopharmaceutical molecules in solution. As a result, the molecular morphology of formulation solution components cannot be fully analyzed. However, structural visualization technology with strengths in visualizing the shape of molecules in solution and analyzing molecular fluctuations would enable the evaluation of conformational changes and mobility in solution where biopharmaceutical molecules composed of proteins, etc., are actually present.

[0011] The present invention has been made in view of the above circumstances, and has as its object to provide a morphological analysis device, a morphological analysis method, and a morphological analysis program that enable the recognition of the morphology of molecular components.

[0012] (1) In order to achieve the above object, the morphological analysis device of the present invention is a morphological analysis device that analyzes the morphology of molecules in a solution, and is equipped with a reference data memory unit that stores reference data, which is molecular shape data of a pre-specified test molecule; a target data memory unit that stores target data, which is molecular shape data of the test molecule to be analyzed; and a morphological estimation unit that estimates the morphology of molecular components in the target data by identifying the state of physical quantities that constitute the molecular shape of the target data relative to the reference data, and is characterized in that both the reference data and the target data have molecular-level resolution.

[0013] (2) Furthermore, in the morphological analysis device described in (1) above, the morphology of the components is a physical difference that directly results from a chemical difference between the components of the molecule identified in the target data and the molecule identified in the reference data.

[0014] (3) Furthermore, in the morphological analysis device described in (1) above, the morphology of the component is characterized in that the difference is due to at least one of the conformation and mobility of the molecule identified in the target data compared to the molecule identified in the reference data.

[0015] (4) In the morphological analysis device described in (2) or (3) above, the molecules identified in the subject data are originally different from the molecules identified in the reference data.

[0016] (5) Furthermore, in the morphological analysis device described in (2) or (3) above, the molecule identified in the target data is the same as the molecule identified in the reference data but at a different time point.

[0017] (6) In addition, in the morphological analysis device described in any one of (1) to (5) above, the morphological estimation unit is characterized in that it identifies the state of the physical quantities constituting the molecular shape by the level of the physical quantities constituting the molecular shape, or the shape or size of the spread expressed as an aggregate of these quantities.

[0018] (7) Furthermore, in the morphological analysis device described in any one of (1) to (6) above, the morphological estimation unit is characterized in that it uses data on physical quantities constituting molecular shapes obtained by measuring a standard sample with low mobility as the reference data.

[0019] (8) In addition, in the morphological analysis device described in any one of (1) to (6) above, the morphological estimation unit is characterized in that it uses accumulated analysis data as training data as the reference data.

[0020] (9) Furthermore, the morphological analysis device described in any one of (1) to (8) above further comprises an output information generation unit that generates output information so that the morphology of the target data can be recognized, the output information generation unit displays the target data on an output device, and receives input of designation information that designates a component whose morphology is to be estimated for the displayed target data, and the morphology estimation unit estimates the morphology of the designated component based on the designation information.

[0021] (10) Furthermore, in the morphological analysis device described in any of (1) to (9) above, an output information generation unit is further provided that generates output information so that the morphology of the target data can be recognized, and the output information generation unit displays the target data on an output device and further displays a specific mark indicating the estimated morphology superimposed on the target data.

[0022] (11) Furthermore, in the morphological analysis device described in any one of (1) to (10) above, the test molecule is a Y-shaped antibody molecule, and the constituent element is Fab or Fc.

[0023] (12) Furthermore, the morphological analysis device described in (11) above further comprises an output information generation unit that generates output information so that the morphology of the target data can be recognized, and a quality judgment unit that judges whether the test molecule conforms to a quality standard based on the estimated morphology, and the output information generation unit is characterized in that, if the test molecule does not conform to the quality standard, the test molecule's non-conformity to the quality standard is displayed.

[0024] (13) Furthermore, the morphological analysis method of the present invention is a morphological analysis method for analyzing the morphology of molecules in a solution, and includes the steps of reading out reference data, which is molecular shape data of a pre-specified test molecule; reading out target data, which is molecular shape data of the test molecule to be analyzed; and estimating the morphology of molecular components in the target data by identifying the state of physical quantities that constitute the molecular shape of the target data relative to the reference data, wherein both the reference data and the target data have molecular-level resolution.

[0025] (14) Furthermore, the morphology analysis program of the present invention is a morphology analysis program for analyzing the morphology of molecules in a solution, and causes a computer to execute the following processes: reading reference data, which is molecular shape data of a pre-specified test molecule; reading target data, which is molecular shape data of the test molecule to be analyzed; and estimating the morphology of molecular components in the target data by identifying the state of physical quantities that constitute the molecular shape of the target data relative to the reference data; and the reference data and target data both have molecular-level resolution.

[0026] FIG. 1 is a schematic diagram showing a morphological analysis system of the present invention. FIG. 2 is a perspective view showing a small-angle X-ray scattering device. FIG. 3 is a block diagram showing details of a morphological analysis unit. FIG. 4 is a flowchart showing a morphological analysis method. (a) and (b) are diagrams showing a projection image of electron density of target data and identified components, respectively. (a) and (b) are diagrams showing marks (arrows), respectively. (a) to (c) are diagrams showing projection images of electron density and marks (arrows) when an antibody molecule to be analyzed is viewed from mutually perpendicular directions, respectively. (a) and (b) are diagrams showing marks (encircled lines), respectively. FIG. 1 is a schematic diagram showing a manufacturing process of a biopharmaceutical. FIG. 2 is a block diagram showing details of a structure identification unit. FIG. 3 is a flowchart showing a structure identification method. FIG. 4 is a schematic diagram showing a structural model, respectively. (a) and (b) are diagrams showing an ideal distribution and an unanalyzable χ ２ 15(a) and (b) are graphs showing the electron density projection images of the raw antibody molecule and the antibody-drug conjugate, respectively. 15(a) and (b) are graphs showing the electron density projection images of the antibody molecule one day after production and two months after production, respectively. 15(b) shows marks indicating the degree of Fab opening. Each structural model and the Rg(ind)-χ ２ The graph shows the points on the graph. The length of the left and right Fabs of the Y-shaped antibody molecule is expressed as χ ２ 10 is a table showing an example of calculation when weighting is performed according to the

[0027] Next, an embodiment of the present invention will be described with reference to the drawings. To facilitate understanding of the description, the same reference numerals are used to designate the same components in the drawings, and duplicated descriptions will be omitted.

[0028] Furthermore, in the following description of the present invention, electron density is used as a representative physical quantity constituting a molecular shape. However, the electron density distribution may be replaced with a physical quantity constituting a molecular shape, such as a potential map, and electron density data may be replaced with molecular shape data, which is data on the physical quantities constituting a molecular shape. The physical quantities constituting a molecular shape refer to quantitative information corresponding to positions that enable identification of the structure formed by the interaction with radiation. The physical quantities constituting a molecular shape include, for example, an electron density map and a potential map, which is a three-dimensional map representing the electric potential within a molecule. Molecular shape data may refer not only to a single structural model, but also to multiple structural models having distributions, as described below.

[0029] [Principle] When X-rays are irradiated onto a molecule moving freely in a solution, a ring of scattered light is generated instead of a spot. By detecting this scattered light and fitting the measured X-ray scattering profile to an electron density map obtained by recursive phase refinement, or an X-ray scattering profile calculated with a hypothetical coordinate structural model, a structural model obtained by fitting the X-ray scattering profile to the measured X-ray scattering profile, or by selecting a structural model with a high fitting agreement index from multiple structural models that make up an ensemble of these, a structural model of a molecule in solution that has a structure with dynamic fluctuations can be accurately reproduced.

[0030] When using such structural analysis techniques, the distribution of electron density in a molecule in solution is determined. The components of a molecule may be at standard angles relative to each other or may be open at both ends beyond the standard. Comparing these cases indicates differences in their conformations. Alternatively, one component may be fixed while the other is able to oscillate in a certain direction. In such cases, the component is said to be mobile. The electron density of the mobile components of a molecule identified by a structural model is locally small, and its distribution extends across the entire range of motion. In this way, the distribution of electron density can be used to estimate the morphology of the molecular components. The estimated morphology of the molecular components includes not only purely physical differences not due to chemical changes between the compared objects, or physical differences indirectly caused by chemical differences, but also physical differences directly resulting from chemical differences. Details of the morphology of molecular components are discussed below.

[0031] [Morphological Analysis System] FIG. 1 is a schematic diagram showing a morphological analysis system 10. The morphological analysis system 10 includes a small-angle X-ray scattering device 100 and a morphological analysis device 200. The small-angle X-ray scattering device 100 measures a small-angle X-ray scattering profile by irradiating X-rays onto a sample S0 and detecting the small-angle scattered X-rays. The sample S0 is suitable for polymers in solution, particularly biopolymers. The sample S0 is preferably a pharmaceutical molecule, molecular complex, or structure in solution. By using X-ray solution scattering with the morphological analysis system 10, it is possible to visualize structural combinations (ensembles) of biopolymers that cannot be observed in a frozen or crystalline state.

[0032] The morphological analysis device 200 is composed of a computer 205, an input device 280, and an output device 290, and controls the operation of the small-angle X-ray scattering device 100, and also acquires and processes measurement data from the small-angle X-ray scattering device 100.

[0033] The small-angle X-ray scattering device 100 includes an X-ray generation unit 110, a sample loading mechanism 120, a detector 130, and a control unit 140. The X-ray generation unit 110 has an X-ray source 111 and irradiates X-rays onto a sample S0. The sample loading mechanism 120 sends a polymer or polymer-free solvent in a solution, which is the sample, to an X-ray irradiation position. The detector 130 detects X-rays scattered by the sample S0 and transmits the obtained measurement data to a computer 205.

[0034] The computer 205 is, for example, a PC, and is configured with a processor that executes processing, and a memory or hard disk that stores programs and data. The computer 205 receives user input from an input device 280 such as a keyboard or a mouse. Meanwhile, the computer 205 displays plots, visualized polymer images, an input screen, and the like on an output device 290 such as a display. The computer 205 may be a server device located on the cloud. Furthermore, in terms of processing load, the function of controlling the operation of the small-angle X-ray scattering instrument 100 and the function of processing measurement data may be separated, with the control being performed by a PC installed on-site and the data processing being performed by a server device.

[0035] [Small-Angle X-Ray Scattering Apparatus] Figure 2 is a perspective view of the small-angle X-ray scattering apparatus 100. The small-angle X-ray scattering apparatus 100 includes an X-ray source 111, an optical system 115, a Kratsky block 117, a sample holder 125, and a detector 130. The X-ray source 111 is a line radiation source or a point radiation source and emits a diverging beam. The optical system 115 is, for example, a KB parallel or serial optical system. The pair of Kratsky blocks 117 interact with the X-rays through their respective edges to define one side and the other side of the X-ray beam. This allows parasitic scattering to be removed from the irradiated X-rays. The sample holder 125 delivers and holds a solution sample of 5 to 10 μl. The detector 130 detects X-rays scattered by the solution sample.

[0036] [Morphological analysis device] Figure 3 is a block diagram showing details of the morphological analysis unit. The morphological analysis device 200 analyzes the morphology of polymers in a solution whose structure has been identified. The functions of the morphological analysis device 200 are mainly realized by a computer 205. The computer 205 comprises a basic function unit 210, a structure identification unit 220, and a morphological analysis unit 230. The components constituting each unit can send and receive information to and from each other via a control bus L.

[0037] (Basic Functional Unit) The basic functional unit 210 performs basic functions such as input / output to and from the user and measurement control. The basic functional unit 210 includes an input / output control unit 211, a measurement control unit 215, and a measurement data storage unit 217. Note that each storage unit within the morphological analysis device 200, such as the measurement data storage unit 217, is merely functionally divided into multiple units, and may physically correspond to either the same memory or multiple memories.

[0038] The input / output control unit 211 controls input and output to the outside. Specifically, it accepts input from the input device 280 and controls output to the output device 290. The input / output control unit 211 can, for example, accept input of measurement conditions and input of conditions for generating multiple structural models. It can also output various plots and the structure of identified polymers.

[0039] The input / output control unit 211 displays the information generated for output on the output device 290. The information generated for output includes the electron density of the test molecule obtained by measurement. The input / output control unit 211 receives input of designation information that designates the constituent element whose morphology is to be analyzed for the displayed target data.

[0040] The measurement control unit 215 controls the operation of the small-angle X-ray scattering device 100. Control includes sending out the sample, generating X-rays, and moving the sample position and detector. Control instructions are sent to a control unit 140 in the small-angle X-ray scattering device 100, which controls each part of the small-angle X-ray scattering device 100.

[0041] The measurement data storage unit 217 stores measurement data of the small-angle X-ray scattering profile detected by the small-angle X-ray scattering device 100. The stored measurement data is used to generate a structural model, calculate indices, and calculate the actual measured value of the molecular size of a polymer.

[0042] (Structure Identification Unit) The structure identification unit 220 generates multiple structure models from the measured X-ray scattering profile obtained by measuring a sample. Then, it calculates the degree of agreement between the calculated X-ray scattering profile calculated from each of the multiple structure models and the measured X-ray scattering profile, and selects a representative structure model from the multiple structure models based on the degree of agreement. In this way, the structure identification unit 220 accurately reproduces a structural model of a polymer in solution that has a dynamically fluctuating structure. Details of the structure identification unit 220 will be described later.

[0043] (Morphological analysis unit) The morphological analysis unit 230 includes a target data extraction unit 231, a reference data storage unit 234, a target data storage unit 235, an output information generation unit 236, a morphological estimation unit 237, and a quality determination unit 238. The morphological analysis unit 230 analyzes the morphology of molecules in solution from the electron density of the target data obtained by the structure identification unit 220.

[0044] The target data extraction unit 231 extracts target data for analyzing the morphology of components from among the structural models identified by the structure identification unit 220. The target data is generated by the structure identification unit 220 selecting a representative structural model from multiple structural models. The electron density data that can be the target data may be one representative structural model or multiple representative structural models with a distribution. The distribution is determined by objective evaluation without arbitrariness, χ ２ Alternatively, multiple structural models may be weighted according to the value of χ ２ It may be a plurality of equivalent structural models in a predetermined range from 1 to .

[0045] For example, χ ２ The weights can be defined so that the closer to 1 the parameter is, the larger the value becomes, and each parameter can be expressed as a weighted average. ２The following formula (1) or (2) can be used as the weighting function, which is a function that monotonically decreases with respect to the absolute value of the difference between a and 1. In either case, a is a real number (usually 0), b and c are positive real numbers, and n is the base of the logarithm.

[0046] In the above example, χ ２ The reference value of is 1, but if the molecule being analyzed takes two states and causes double dispersion, the reference value may deviate from 1. In that case, the offset is taken into account when calculating the reference value of χ ２ The weighting is defined so that the closer to the reference value, the larger the value. For example, when analyzing PCT characteristics, an offset (a) may be necessary when two components coexist.

[0047] The reference data storage unit 234 stores reference data. Reference data is electron density data of a test molecule that has been specified in advance. The reference data is, for example, registered as electron density data of a known test molecule. The reference data is preferably analysis data that has been previously analyzed and accumulated by the morphological analysis device 200, but may also be data composed of known information. "Previous" means before the morphological analysis currently being performed. The electron density data that can serve as reference data may be a single representative structural model, as in the case of the target data, or multiple representative structural models with a distribution.

[0048] The target data storage unit 235 stores target data representing the electron density of a test molecule. Target data is electron density data of the test molecule to be analyzed. Both the reference data and the target data have molecular-level resolution. "Molecular-level resolution" is preferably a resolution of 3 Å to 10 Å, for example, for a size of 140 Å to 350 Å of the constituent elements of the test molecule whose morphology is to be analyzed. Note that while the target data is data obtained by measurement, the reference data does not necessarily have to be obtained by measurement. The reference data may be data calculated from a known crystal structure, for example.

[0049] The output information generating unit 236 generates output information so that the form of the target data can be recognized. Preferably, the output information generating unit 236 generates output information based on the estimated form of the test molecule. The output information generating unit 236 converts, for example, the electron density distribution projected in a predetermined direction into shades of a single color or shades of warm and cool colors according to the position. For example, positions with high electron density can be represented by a dark single color or a dark warm color. Positions with low electron density can be represented by a light single color or a light warm color. The electron density distribution can also be represented by a two-color gradation, such as a blue-to-red gradation. This allows differences in the components of the test molecule resulting from at least one of the conformation and mobility to be represented.

[0050] The output information generating unit 236 may generate output information that visually displays the difference between the target data and the reference data, without being based on the estimated morphology of the test molecule. For example, the output information may be generated to display the target data superimposed on the reference data.

[0051] The output information generating unit 236 extracts the states and conditions associated with the test molecule and displays these states and conditions together with the electron density distribution of the test molecule. For example, the states and conditions of an antibody molecule include the presence or absence of a sugar chain, changes over time, culture conditions, nutrient conditions, extraction method, and transportation method.

[0052] The output information generating unit 236 can also display marks along with the electron density of the test molecule as a result of analyzing the morphology of the constituent elements. The marks are preferably displayed superimposed on the display of the electron density of the constituent elements. This allows the parameters of the molecular morphology to be displayed in a visually easy-to-understand manner. For example, the degree of opening of the Y-shape can be used to visualize conformational characteristics, and a line segment with arrows at both ends (arrow mark) can be used to display the direction and distance of motility. For example, it is also possible to display the propagation of the movement of the test molecule with an arrow.

[0053] Additionally, it is possible to mark voids and low-density areas with balloon-shaped encircling lines (encircling line marks). In this way, it is effective to use marks that can indicate the size of the space for voids, etc. This encircling line mark is a label that indicates the molecular structure, and represents physical differences that directly result from chemical differences in the constituent elements that induce voids, etc. Furthermore, in the case of changes over time, it is also possible to mark areas of increased mobility.

[0054] If the morphologically analyzed test molecule does not conform to the quality standard, the test molecule may be displayed as not conforming to the quality standard, for example, if parameters representing the morphology of the molecular components do not fall within a predetermined range of values.

[0055] The morphology estimation unit 237 reads the reference data and the target data, and estimates the morphology of the molecular components by identifying the state of the electron density of the target data relative to the reference data. This allows for the recognition of differences in the morphology of the molecular components, which are attributable to at least one of conformation and mobility. As a result, for example, in the manufacturing process of biopharmaceuticals, differences attributable to at least one of the conformation and mobility of the evaluated molecular structure can be used for quality control.

[0056] The morphology estimation unit 237 identifies the state of the electron density distribution based on the level of electron density, the shape of the spread, or the size. This allows the morphology of the molecular components to be recognized, and as a result, the morphology of the test molecule to be estimated. Note that size includes, for example, cross-sectional area and distance.

[0057] The morphology estimation unit 237 can identify, for example, each Fab or Fc of an antibody molecule as the components of the test molecule. For example, each component can be identified by determining the center position of the test molecule and determining the edges based on that center position. The components can be considered as ellipsoids, and the length of the minor axis including the center can be calculated as the size of the component. The degree of opening can also be determined by identifying the central axes of each component and determining the angles they form. Such morphology information can be calculated as the difference or ratio of the target data to the reference data.

[0058] Conformation is the spatial arrangement of each domain in a molecule. Spatial arrangement is related to the positional information of the electron density and includes the relative positions and angles of the components. A change in conformation refers to a change in the spatial arrangement of a domain at one measurement time to the spatial arrangement of the domain at another measurement time. Motility is related to the level of electron density and is the movement of mobile components within a molecule. Motility refers to a state in which a molecule moves without converging to a predetermined arrangement within a certain measurement time range. Motility includes the speed, distance, range, and direction of movement.

[0059] The morphology estimation unit 237 may evaluate structural changes in the test molecule, identify changes over time, evaluate flexibility, or identify flexible regions based on differences resulting from at least one of the conformation and mobility of the evaluated test molecule. For example, it may identify differences in the degree of opening and flexibility of each Fab of an antibody molecule. The morphology estimation unit 237 may also evaluate abnormalities in the entire molecule as damage to the structure.

[0060] The morphology estimation unit 237 can use electron density data obtained by measuring a standard sample with low mobility as reference data, thereby making it possible to evaluate the morphology of molecular components using the measurement results of the standard sample even if known electron density data is not available.

[0061] The morphology estimation unit 237 may have an AI function, such as a machine learning model. When the morphology estimation unit 237 has an AI function, the accumulated analytical data can be used as reference data and as training data. This makes it possible to use the accumulated analytical data as training data to analyze and evaluate the morphology of the components of the test molecule in the target data using AI, even when there is no standard sample or clear reference data.

[0062] For example, a machine learning model consisting of a neural network with an input layer, intermediate layer, and output layer can be used. When the electron density distribution of a test molecule and its morphological or quality information are known, these can be associated and used as training data to train the machine learning model. That is, the electron density distribution of the test molecule is input, and the weights of each neuron in the intermediate layer are tuned by repeated training so that the output layer outputs correct morphological or quality information with a high probability. A convolutional neural network may be used in this case to roughly recognize the target. This allows the morphology of the test molecule to be estimated and its quality to be determined simply by inputting the electron density distribution of the target data into the machine learning model, without the need for special parameters.

[0063] The morphology estimation unit 237 preferably evaluates differences resulting from at least one of the conformation and the motility designated based on the designation information, thereby obtaining evaluation results for components of the test molecule for which it is desired to evaluate differences resulting from at least one of the conformation and the motility designated by, for example, a range designation using a mouse.

[0064] The morphology estimation unit 237 may compare the molecular structure or mobility evaluated at a certain time t1 for the same sample with the molecular structure or mobility at a certain time t2 (t1<t2) after a certain time has elapsed. By clarifying the difference resulting from at least one of the conformation and mobility due to the change over time obtained in this way, the user can confirm the influence of time.

[0065] The morphology estimation unit 237 can also evaluate the stability of the test molecule. For example, the morphology estimation unit 237 can determine the presence or absence of components of the test molecule. If a component is absent, it is preferable to also estimate the position and size of the void. For example, it can evaluate the stability of the Fc of an antibody molecule due to glycosylation. It can also determine the loss of the glycosylation originally bound to the Fc of the antibody molecule. The morphology estimation unit 237 can also evaluate the non-equivalence of components. For example, it is possible to evaluate the non-equivalence of the Fab of an antibody molecule. In this way, if there is a region where the motility is increased due to the state of the antibody molecule being different from the reference state, that region can be identified and displayed.

[0066] The quality assessment unit 238 determines whether the test molecule conforms to quality standards based on the estimated morphology of the test molecule. For example, for a Y-shaped antibody molecule, it determines whether the parameters indicated by the estimated morphology fall within a predetermined value range. Examples of parameters indicated by the morphology include numerical values ​​indicating the size, angle, or mobility of the constituent elements. For example, this corresponds to whether the degree of opening of the Fab, which estimates the conformational difference, falls within a predetermined value range. The predetermined value is a standard threshold based on reference data, or a threshold including a confidence limit. For example, in the manufacturing process of a Y-shaped antibody molecule, an antibody whose left and right Fabs are open by an angle of 140° or more, which is the predetermined value, is considered to be a deteriorated antibody.

[0067] If the parameter does not fall within a predetermined value range, the quality determination unit 238 can identify the product as defective. For example, the test molecule is a Y-shaped antibody molecule, and the constituent elements are Fab or Fc. By performing measurement and analysis under such settings, the quality of the product can be determined in the manufacturing process of the Y-shaped antibody molecule. Note that, instead of using predetermined values, information corresponding to the morphology of the test molecule and information on quality standards may be input into an AI learning model to determine the suitability of the quality.

[0068] [Morphological Analysis Method] The operation of the morphological analysis device 200 configured as above will now be described. Fig. 4 is a flowchart showing the morphological analysis method for motility. First, the morphological analysis device 200 reads out the reference data and the component identification information (step S1).

[0069] The reference data is, for example, known electron density data of an antibody molecule whose morphology is to be analyzed. Morphological analysis includes evaluation of differences resulting from at least one of conformation and mobility. The component-specific information is information for identifying portions corresponding to each component from the electron density data of the test molecule. The read information may further include information specifying in advance the component whose morphology is to be analyzed.

[0070] Next, the target data is read (step S2) and displayed (step S3). The target data is displayed as a projected image in a predetermined direction. The morphological analysis device 200 then accepts an input specifying the component of the test molecule whose morphology is to be analyzed (step S4). For example, an example of a component of an antibody molecule is Fab. If no specification is required, the specification may be skipped.

[0071] The morphological analysis device 200 determines whether there is information specifying a component (step S5), and if no component is specified, identifies all components (step S6). On the other hand, if there is information specifying a component, it identifies the specified component (step S7). For example, it assigns the Fc, left Fab, and right Fab of an antibody molecule as components. Details of identifying the components will be described later.

[0072] The morphology analysis device 200 then estimates the morphology of the identified component using the state of electron density (step S8). For example, the electron density distribution representing the Fab of an antibody molecule is sliced ​​along a plane including the central axis of the Fab, and the vertical and horizontal widths of the cross section are detected. The vertical is the direction of the central axis, and the horizontal is the direction perpendicular to the vertical. Furthermore, the cross-sectional area of ​​the component is roughly calculated from the vertical and horizontal widths of the cross section. The vertical and horizontal widths can be applied to the major and minor axes to calculate the area of ​​an ellipse.

[0073] The morphological analysis device 200 may display a mark representing the morphology according to the estimated morphology (step S9). Marks representing motility include, for example, an arrow mark. Marks representing conformational changes include, for example, an encircling line mark. Details of the various marks will be described later.

[0074] The morphology analysis device 200 judges the morphology estimation result (step S10). For example, for the Fab of an antibody molecule, if its cross-sectional area is larger than a standard, it is judged to have high motility, and if it is smaller, it is judged to have low motility. Furthermore, based on the judgment, it may be possible to confirm, for example, that the sample is not defective. Then, the judgment result is displayed (step S11), and the series of operations is completed.

[0075] When evaluating motility, the width of the cross section of the left and right Fabs of the antibody molecule may be compared, and the Fab with the longer width may be evaluated as a Fab with high motility, and the Fab with the shorter width may be evaluated as a Fab with low motility. Then, it may be determined whether there is a difference in motility between the two Fabs.

[0076] Furthermore, for example, the orientation (degree of lift) of the Fabs forming the Y-shape in a Y-shaped antibody molecule can also be evaluated. In this case, the evaluation may be performed as the angle of the central axis of each Fab relative to the central axis of the Fc, or as the angle between the Fabs. If the angle of the Fab relative to the Fc is equal to or less than a predetermined value, or if the angle between the Fabs is equal to or greater than a predetermined value, it is determined that the Fabs have lowered from their original positions and the antibody molecule has become closer to a T-shape, and the sample can be deemed defective. In this way, the degree of opening of the Y-shape of the antibody molecule can be used as an indicator for evaluating the freshness of the sample.

[0077] [Identification of Components] Figures 5(a) and (b) are diagrams showing a projection image of the electron density of the target data and the identified components, respectively. Figure 5(a) is a projection image showing the electron density distribution of the target data of a human serum antibody molecule 401. The projection image is an image in which the electron density of the human serum antibody molecule is projected in a certain direction. The human serum antibody molecule 401 shown in the projection image maintains a Y-shape.

[0078] The Fab and Fc regions are detected as components of this electron density distribution using a Y-shaped search. Figure 5(b) shows the detection results for each region. As shown in Figure 5(b), the Fab and Fc regions located vertically above and below in the solution are detected, and marks 500a, 500b, and 500c (line segments) indicating the left Fab, right Fab, and Fc regions, respectively, are displayed. Since the projection direction is fixed, it is possible to identify the left Fab and right Fab.

[0079] [Morphology of Components] (Observation of Physical Differences Directly Resulting from Chemical Differences) Comparing reference data with target data allows us to estimate the morphology of the components of the target data. The morphology of components can be expressed as physical differences resulting from chemical differences between the components of the molecules identified in the reference data and those identified in the target data. These chemical differences can include short-circuiting, deletion, elongation, or the addition of some substance to the structure of the molecule identified in the target data compared to the reference molecule. If these chemical changes are significant, short-circuiting and deletion will manifest as a lack of electron density in the relevant area, while extension and the addition of some substance will manifest as excess electron density. For example, when developing new designs such as bispecific antibodies or culturing conditions, quickly understanding the state of glycans or unintended chemical modifications can help determine success or failure and improve process efficiency. Such physical differences directly resulting from chemical differences primarily occur when comparing different targets.

[0080] (Observation of physical differences regardless of chemical differences) The morphology of a component may be a difference resulting from at least one of the conformation and mobility of a molecule identified in the subject data relative to a molecule identified in the reference data. In this case, even if the difference in chemical structure, such as a defect, is extremely small and not observed as a direct physical difference such as a defect in electron density, it may result from an indirect influence resulting from at least one of the conformation and mobility. In this case, the molecule identified in the subject data may be originally different from the molecule identified in the reference data, or they may be the same entity, just compared at different times.

[0081] The difference resulting from at least one of conformation and motility includes a difference in conformation, a difference in conformation distribution, a difference in motility, or a difference in motility distribution.

[0082] (When the comparison targets are originally different) When the comparison targets are originally different, it is effective to look at differences arising from at least one of conformation and mobility. For example, this can be done when using a molecule identified in reference data as a standard and observing a molecule identified in the target data. Examples of situations include designing antibody drug functions for the same antibody, or analyzing molecules that produce different functions even though the same antibody drug has been created as the standard.

[0083] (When comparing the same object at different times) Even when comparing the same object at different times, it is effective to observe differences resulting from at least one of conformation and mobility. For example, it is possible to observe the time-dependent changes of a single molecule. In this case, it is possible to observe changes or distributions in at least one of conformation and mobility. This is effective for the formulation development of antibody drugs and the quality control of manufactured antibody drugs.

[0084] [Mark Display] The morphology estimated using the electron density state of the constituent elements can be represented using marks. In the example shown in FIG. 5(b), the left Fab, right Fab, and Fc constituent elements are represented by marks 500a, 500b, and 500c (line segments), respectively. In this way, the marks can indicate the conformation of the test molecule. The marks 500a, 500b, and 500c can also be represented by double lines along the molecular chain.

[0085] The marks can also indicate the motility of the test molecule. Figures 6(a) and (b) are diagrams showing marks (arrows) 501 and 502, respectively. In the example shown in Figure 6(a), a double-headed line segment indicating the motility of the left Fab of human serum antibody molecule 402 is displayed superimposed on a projection image of the electron density of the target data obtained for human serum antibody molecule 402. The length of the line segment corresponding to the width of the cross section including the central axis of the Fab represents the magnitude of motility, and the arrows at both ends can indicate the directionality of the movement. This mark 501 visualizes that the left Fab is narrow in width and has no spatial extent, clearly indicating its low motility.

[0086] 6(b), a line segment with arrows at both ends indicating the mobility of the right Fab of human serum antibody molecule 403 is displayed superimposed on the electron density projection image of the target data obtained for human serum antibody molecule 403. This mark 502 indicates that the right Fab is significantly wider than the left Fab, visualizing its spatial extent and clearly indicating its high mobility.

[0087] The marks indicating the evaluated motility as described above are three-dimensional vector quantities, and their shapes on the screen vary depending on the projection direction of the antibody molecule being analyzed. Figures 7(a) to 7(c) are diagrams showing electron density projection images and marks (arrows) when the antibody molecule 404 being analyzed is viewed from mutually perpendicular directions. Figures 7(b) and 7(c) are diagrams showing the antibody molecule 404 and mark 503 as viewed from directions 7B and 7C shown in Figure 7(a) (right side view and plan view when Figure 7(a) is considered a front view). This indicates that the Fab region in particular appears to be more spread out in the projection image of Figure 7(c), indicating that it is moving in a direction perpendicular to the plane formed by the Y-shape.

[0088] Marks can also be used to indicate physical differences that result directly from chemical differences in the molecular components. Examples of physical differences that result directly from chemical differences in the molecular components include cleavage or loss of components. Figures 8(a) and 8(b) are diagrams showing marks (encircled lines) 504 and 505, respectively. In the example shown in Figure 8(a), the glycan attached to the Fc in the antibody molecule 405 is cleaved, resulting in the loss of the central glycan in the Fc and cleavage of the Fc. If the structure of the control data contains a defect or other defect compared to the structure of the reference data, the defect area can be indicated by an encircled line. In the example shown in Figure 8(a), a balloon-shaped encircled line mark 504 is displayed superimposed on the cleaved portion of the Fc, indicating the position and size of the cleaved portion.

[0089] In the example shown in Figure 1, it can be seen that a void has occurred in the center of the Fc in antibody molecule 406. This void suggests the possibility that there may be no glycans originally bound to the Fc. The position and size of the void can be indicated by displaying a balloon-shaped encircling line mark 505 superimposed on the void. Note that the mark is not limited to an arrow or an encircling line, and may be a geometric representation such as a line, a curve, or a point (including the start point and end point) that allows recognition of physical differences that directly result from chemical differences in the constituent elements of the molecule.

[0090] [Application to Pharmaceutical Manufacturing Processes] The above-described method for analyzing the morphology of test molecules is particularly suitable for application to the manufacturing process of biopharmaceuticals, which contain proteins as components. Figure 9 is a schematic diagram showing the manufacturing process of biopharmaceuticals. Biopharmaceuticals are produced by screening antibody molecules or vectors cultured under specified conditions, selecting and purifying the resulting culture strains, and culturing the desired biopharmaceuticals in large quantities.

[0091] In this case, factors that determine the conditions for culturing and purifying raw materials include structural identity (LC, LC / MS), glycan structure as a characteristic analysis, its heterogeneity (correlation with activity), and vector completion (stability of the viral capsid), and solutions that do not meet a certain quality are eliminated.In this way, the concept of "quality by design (QbD)" is important, in which quality is designed while constantly monitoring and optimizing the manufacturing process.

[0092] The above-mentioned method for analyzing the morphology of test molecules is effective in determining whether a certain level of quality is met. Conventional structural analysis methods require extensive time and effort for prior method development, but this method allows even inexperienced users to perform characteristic analysis quickly and easily. It also reduces the costs and human burden associated with process optimization in discovery research and quality control.

[0093] In the above example, the morphological analysis method is mainly applied to antibody molecules, but it may also be applied to molecules that store and transport substances. Ferritin molecules have a cavity inside, and the hatch portion opens and closes to store the substance to be transported inside. For example, the above morphological analysis method can evaluate the openability of the hatch portion of a capsid molecule and confirm the integrity of its shape.

[0094] 10 is a block diagram showing the details of the structure identification unit. The morphology analysis device 200 acquires data of the X-ray scattering profile measured by the small-angle X-ray scattering device 100, and identifies the structure of the polymer based on the data.

[0095] The structure identification unit 220 includes a structural model generation unit 221, a theoretical scattering intensity calculation unit 222, an index calculation unit 223, a correlation determination unit 224, a trend analysis unit 225, a theoretical size calculation unit 226, an actual size calculation unit 227, a comprehensive determination unit 228, and a structural model selection unit 229.

[0096] The structural model generating unit 221 generates a plurality of structural models from the actually measured X-ray scattering profile obtained by measuring the sample S0. The details of the generation of the structural models will be described later.

[0097] From the obtained structural models, an index χ was calculated, which indicates the degree of agreement between the measured data and the calculated scattering curve for each structural model. ２ and the calculated value Rg(ind) of the dynamic radius of gyration of the electron density model are calculated and plotted to create a first plot (see FIG. 13 described later). The dynamic radius of gyration Rg of a molecule is a suitable example of a parameter representing the molecular size, and other parameters representing molecular sizes may also be used.

[0098] It is preferable that the structural model generation unit 221 generates multiple structural models by changing conditions when there is a trend in the distribution of the first plot. This makes it possible to attempt to regenerate a structural model when none of the structural models are valid. The structural model generation unit 221 can also generate multiple structural models under conditions based on user instructions. This makes it possible to attempt to generate a structural model by changing conditions when none of the structural models are valid.

[0099] The structural model generation unit 221 can generate each of the multiple structural models one by one through repeated processing according to settings. This makes it possible to allocate necessary computational resources and proceed with processing while reducing unnecessary processing and confirming the validity of the structural models. The structural model generation unit 221 may also generate each of the multiple structural models at once through parallel processing according to settings. This makes it possible to reproduce a highly valid structural model of a biopolymer in a short period of time. The theoretical scattering intensity calculation unit 222 calculates the theoretical scattering intensity from each of the multiple structural models.

[0100] The index calculation unit 223 calculates an index that indicates the degree of agreement between the calculated X-ray scattering profile calculated from each of the plurality of structural models and the actually measured X-ray scattering profile. Specifically, the index that has been subjected to statistical processing is χ ２ However, there are no particular limitations as long as it is an index that indicates the degree of agreement with the X-ray scattering profile. ２ Other than χ, there are parameters of normal distribution or Poisson distribution, R value, RMS value and RMD value, which are indices that indicate the degree of agreement between the structure factor calculated from the measured diffraction data and the structure factor based on the model structure obtained by analysis. ２ can be calculated as follows:

[0101] The correlation determination unit 224 creates a first plot by plotting the parameters representing the molecular size against the calculated index for each of the plurality of structural models, and determines whether or not there is a correlation in the first plot. The correlation determination unit 224 preferably displays the first plot on a display. This allows the user to visually check whether or not there is a correlation in the first plot.

[0102] When performing repeated processing, the correlation determination unit 224 preferably determines whether or not there is a correlation between the plots for each iteration. This allows the processing to proceed while checking the validity of the structural model for each iteration. As a result, efficient processing is possible when computational resources are limited.

[0103] When there is no correlation in the first plot, the trend analysis unit 225 performs multivariate analysis on the distribution of the first plot. Examples of multivariate analysis include PCA (principal component analysis). This makes it possible to determine the presence or absence of a trend even when there is no correlation. For example, even when there is no correlation in the overall data, by separating multiple types of correlated data, each type of data can be used as correlated data.

[0104] The theoretical size calculation unit 226 creates a second plot by plotting parameters representing molecular sizes calculated based on multiple structural models for each voxel size. The calculated molecular size parameters are χ of the regression line of the calculated radius of gyration Rg(ind). ２ Preferably, the intercept Rg(sect) at ρ = 1. Then, the theoretical size calculation unit 226 uses the second plot to calculate a calculated value Rg(calc) of a parameter that represents the molecular size of a polymer in a solution independent of the voxel size.

[0105] The actual size calculation unit 227 calculates a parameter representing the molecular size of the polymer in solution as an actual value from the measured X-ray scattering profile. Specifically, a Guinier plot is performed to calculate the actual value Rg(exp), which is the dynamic radius of gyration of the molecule. Note that if the sample forms a hydration sphere in the solution, it is preferable to treat the value obtained from the Guinier plot minus a predetermined value for the hydration sphere as the actual value Rg(exp).

[0106] The comprehensive determination unit 228 determines whether there is a statistically significant difference between the calculated value Rg(calc) of the radius of gyration calculated by the theoretical size calculation unit 226 and the actual measurement value Rg(exp) calculated by the actual size calculation unit 227. If it is determined that there is no significant difference, a representative structural model is selected from the multiple structural models. If it is determined that there is a significant difference, the process is terminated.

[0107] The structural model selection unit 229 selects a representative structural model from among the plurality of structural models based on the calculated index. ２ A structural model whose σ is close to 1 is selected as the representative structural model. Because the representative structural model is selected based on the index, it is possible to accurately reproduce the structural model of a polymer in solution that has a dynamically fluctuating structure. As a result, it is possible to accurately visualize even biopolymers in solution for which there is no prior information.

[0108] It is preferable that the structural model selection unit 229 does not select a representative structural model if there is no correlation in the first plot. This makes it possible to stop identifying a structural model and avoid unnecessary calculations if none of the structural models are valid. Then, it is possible to regenerate a structural model or perform a re-experiment depending on the situation.

[0109] It is preferable that the structural model selection unit 229 does not select a representative structural model unless the difference between the calculated value and the measured value is within a predetermined range. This makes it possible to stop identifying a structural model if the validity of the structural model is not guaranteed in terms of molecular size, even if the structural model can be identified.

[0110] [Structure Identification Method] (Overall Method) A method for identifying the structure of a polymer in a solution using the morphology analysis system 10 configured as described above will be described. FIG. 11 is a flowchart showing the structure identification method. First, the small-angle X-ray scattering device 100 sends the sample S0 in the solution to a predetermined position and irradiates the sample S0 with X-rays. The small-angle X-ray scattering device 100 detects the scattered X-rays and transmits them to the computer 205 as small-angle X-ray scattering profile data. The computer 205 stores the received X-ray scattering profile data.

[0111] In response to a user's instruction, the computer 205 reads out data of an X-ray scattering profile obtained from the sample S0 whose structure is to be identified (step S1), and acquires the structural model generation conditions, such as the box size, voxel size, and number of trials for each voxel size, specified by the user (step S2).

[0112] A structural model is generated from the read X-ray scattering profile in accordance with the acquired generation conditions (step S3). The details of the generation of the structural model will be described later. Next, a theoretical scattering intensity is calculated based on the generated structural model (step S4). An index χ, which represents the degree of agreement between the calculated X-ray scattering profile and the measured X-ray scattering profile, is calculated based on the read measured X-ray scattering profile and the calculated theoretical scattering intensity. ２ and the calculated value Rg(ind) of the radius of gyration of the particle for each structural model is calculated (step S5), and χ２ A plot of −Rg(ind) is created (step S6).

[0113] Next, as a repetition condition, it is determined whether or not a predetermined number of attempts have been made to generate a structural model for a specific voxel size (step S7). If it is determined that the predetermined number of attempts have not been made, the process returns to step S3. If it is determined that the predetermined number of attempts have been made, the process proceeds to step S8. The predetermined number of attempts is, for example, 50 times. Note that, although the process is repeated for a specific voxel size in step S7 in the above example, the process may simply be repeated a predetermined number of times.

[0114] The χ thus created for a particular voxel size ２ It is determined whether or not there is a correlation in the plot of -Rg(ind) (step S8). Details of the correlation determination process will be described later. If it is determined that there is no correlation, it is determined whether or not there is any trend in the plot using multivariate analysis or the like (step S9). If it is determined that there is a trend, the process returns to step S3, the conditions are changed, and a structural model is created again. If it is determined that there is no trend, the series of processes ends without selecting a representative structural model. By determining whether or not there is a correlation at the stage when the repeated process of step S7 is completed and terminating processes that are unlikely to occur, computational resources can be used efficiently.

[0115] On the other hand, if it is determined in step S8 that there is a correlation, it is determined whether the condition for ending the repetition, that is, completing the generation of structural models for all of the multiple voxel sizes, is met (step S10).If it is determined that the condition is not met, the voxel size is changed and the process returns to step S3.

[0116] In step S10, if the condition for ending the repetition is met, χ ２ - A regression line is obtained from the plot of Rg(ind), and the χ of the regression line is ２The intercept Rg(sect) at s = 1 is calculated (step S11). Then, a plot of the intercept Rg(sect) against the voxel size is created (step S12), a regression line is obtained from the plot, and the calculated value Rg(calc) of the molecular radius of gyration is calculated as the extrapolated value where the voxel size is zero on the regression line (step S13). Meanwhile, a Guinier plot is created for the read X-ray scattering profile, and the measured value Rg(exp) of the molecular radius of gyration is calculated (step S14).

[0117] Next, the validity of the calculated radius of gyration Rg(calc) is determined by determining whether the difference between the calculated radius of gyration Rg(calc) and the measured radius of gyration Rg(exp) falls within a certain range (step S15). In this case, if the sample forms a hydration sphere in the solution, it is preferable to treat the measured radius of gyration Rg(exp) as the value obtained by subtracting a predetermined value for the hydration sphere from the value obtained from the Guinier plot. If a hydration sphere is formed, the predetermined value is preferably 1.5 Å to 2.0 Å.

[0118] In step S15, if it is determined that the calculated value Rg(calc) is not valid, the series of processes is terminated without selecting a representative of the structural model. ２ A representative structural model is selected from those for which the value is closest to 1, and the selected structural model is output to the output device 290 (step S16), thereby completing the series of processes.

[0119] The representative structural model does not necessarily have to be single, but may be multiple. For example, χ ２ If there is a correlation for each voxel size in the plot of −Rg(ind), there may be as many representative structural models as the number of correlations as an ensemble. In the above process, the same number of representative structural models as the number of correlations may be selected.

[0120] In the above example, multiple structural models are generated for each iteration, but they may be generated all at once using parallel processing. Alternatively, a fixed number of parallel processes may be performed repeatedly. The process to be selected can be determined based on which is more important: computational resources or the speed at which results are obtained.

[0121] (Generation of Structural Model) Next, the generation of the structural model will be described in detail. When a biopolymer in a solution is irradiated with X-rays, the structure of the biopolymer is irradiated with X-rays. －１ A gentle ring-shaped scattered X-ray intensity peak appears in the range of about 0.7 Å. By integrating this in the circumferential direction, a small-angle X-ray scattering profile can be obtained. －１ By acquiring and analyzing scattering intensity data with high precision between 1 and 2, it becomes possible to directly visualize electron density that is meaningful for the actual structure.

[0122] Figure 12 is a schematic diagram showing a structural model. To generate a structural model, the volume of a cubic box with a side length of H in real space containing a polymer is first discretized into an N x N x N grid of cubic voxels (N = 4 in the example shown in Figure 12). As the shading of each voxel in Figure 12 indicates, the density f(x, y, z) of each voxel is randomly assigned a value within a certain range. Then, the three-dimensional reciprocal space intensity is calculated from the three-dimensional structure factor, and the structure is divided into concentric shells as a function of the magnitude of the scattering vector q.

[0123] The three-dimensional scattering intensity is then converted to a one-dimensional profile and compared with the experimental scattering data. The three-dimensional structure factor is scaled to match the experimental data for each q concentric shell, and a new electron density map is created in real space by inverse Fourier transformation. The density outside the map is set to zero. A new structure factor is obtained by forward Fourier transformation, and this cycle is repeated until convergence. In this way, a different structural model is generated for each trial, and the index χ is calculated for each structural model. ２ and the calculated value Rg(ind) is calculated.

[0124] (Correlation determination process) χ calculated for each structural model ２By plotting Rg(ind) and Rg(ind) and determining whether they are correlated with each other, the validity of the structural model for the measured data can be determined. Whether the plot is correlated can be objectively determined, for example, by using a correlation coefficient. Furthermore, if there is no correlation, by determining whether there is a trend in the plot, it can be determined whether the data is unpromising or whether it could be meaningful depending on the measurement conditions.

[0125] 13(a) and (b) show the ideal distribution and the unanalyzable χ ２ 13(a) is a graph showing the distribution of χ −Rg(ind). ２ A correlation appears in the plot of −Rg(ind), and it can be seen that multiple structural models are valid. On the other hand, in the distribution that cannot be analyzed, as shown in Figure 13(b), ２ The plot of -Rg(ind) shows no correlation or trend, indicating that the multiple structural model is not valid. If the structural model is valid, χ ２ A structural model whose σ is within a predetermined range from 1 can be selected as the representative structural model.

[0126] Example 1 In the manufacturing process of an antibody-drug conjugate, the electron density of a raw antibody molecule and an antibody-drug conjugate was observed using the structural model identification method described above. Figures 14(a) and 14(b) are figures showing electron density projection images of a raw antibody molecule and an antibody-drug conjugate, respectively. Both show electron density projection images of a Y-shaped antibody molecule, and it can be seen that a raw antibody molecule with a small electron density spread is transformed into an antibody-drug conjugate with a large electron density spread. The left Fab, right Fab, and Fc regions of the antibody molecule are indicated by line marks 600a to 600c.

[0127] Furthermore, using the structural model identification method described above, the state of the manufactured antibody-drug conjugate was observed one day and two months later. Figures 15(a) and (b) show electron density projection images of the antibody molecule one day and two months after manufacture, respectively. As indicated by the arrows 601 and 602, it can be seen that the Fab motility of the antibody molecule two months later is higher than that of the antibody molecule on day one. Over time, the Fab, which is normally Y-shaped, begins to droop, and the antibody molecule changes from a Y-shape to a significantly more open Y-shape. It can also be seen that the fluctuation of the Fab becomes greater. This is thought to be the result of the degradation of the antibody-drug conjugate after manufacture.

[0128] [Example 2] Electron density data that could be used as target data was calculated for a Y-shaped antibody molecule. χ ２ The parameters were calculated as a weighted average based on the observed values ​​weighted according to the value of . By evaluating each structural model with a predetermined weighting function for all assumed structural models, an objective evaluation without arbitrariness becomes possible. Figure 16 shows each structural model and the Rg(ind)-χ ２ This is a schematic diagram showing points on a graph. The Fab lengths 701a to 705a on the left and 701b to 705b on the right of the Y-shaped antibody molecule represent the Fab lengths before correction in each structural model. ２ The structural model whose σ is closest to 1 is the representative structural model.

[0129] In the example shown in FIG. 16, χ ２ The weighting function used was the formula (1) where a = 0, b = 4, and n = 10. Figure 17 shows the structure of the Y-shaped antibody molecule, where the length of the left and right Fab fragments is expressed as χ ２ This is a table showing an example of calculation when weighting is performed according to the structure model No. 1 to 50. ２The weighted corrected observed values ​​of the left and right Fab lengths were calculated according to the weighted average length and standard deviation σ. The weighted average length and standard deviation σ were calculated by summing the calculated corrected observed values. In this way, it is possible to calculate the parameters of the molecular shape data of the test molecule to be analyzed. In this example, the Fab length was used as the parameter, but other parameters such as the opening angle between the arms may also be used.

[0130] 10 Morphological analysis system 100 Small-angle X-ray scattering device 110 X-ray generation unit 111 X-ray source 115 Optical system 117 Kratsky block 120 Sample loading mechanism 125 Sample holding tube 130 Detector 140 Control unit 200 Morphological analysis device 205 Computer 210 Basic function unit 211 Input / output control unit 215 Measurement control unit 217 Measurement data storage unit 220 Structure identification unit 221 Structural model generation unit 222 Theoretical scattering intensity calculation unit 223 Index calculation unit 224 Correlation determination unit 225 Trend analysis unit 226 Theoretical size calculation unit 227 Actual size calculation unit 228 Overall determination unit 229 Structural model selection unit 230 Morphological analysis unit 234 Reference data storage unit 235 Target data storage unit 236 Output information generation unit 237 Morphology estimation unit 238 Quality judgment unit 280 Input device 290 Output device 401 to 406 Human serum antibody molecules 500a to 500c, 600a to 600c Marks (line segments) 501 to 503, 601, 602 Marks (arrows) 504, 505 Marks (encircling lines) L Control bus

Claims

1. A morphology analysis device that analyzes the morphology of a molecule that is placed in a state where it can exist in multiple molecular shapes, a reference data storage unit for storing reference data, which is molecular shape data of one of the test molecules; a target data storage unit for storing target data, which is molecular shape data of the other test molecule; a shape estimation unit that, on the assumption that the reference data and the target data are each a plurality of possible molecular shape data obtained as an ensemble, identifies parameters indicating shape from distributions of physical quantities, such as electron density or potential maps, that constitute the molecular shapes of the reference data and the target data, and that are obtained with a resolution of 3 Å or more and 10 Å or less, and compares the parameters indicating shape identified from the reference data with the parameters indicating shape identified from the target data; Equipped with the morphological analysis device determines whether or not a molecule identified in the reference data matches a molecule identified in the target data based on a result of the comparison; A morphological analysis device characterized in that both the reference data and the target data have molecular level resolution.

2. 2. The morphology analysis device according to claim 1, wherein the morphology estimation unit weights each of the plurality of molecular shape data and identifies parameters indicating the morphology from the state of physical quantities constituting the molecular shapes of the reference data and the target data.

3. 3. The morphological analysis apparatus according to claim 1, further comprising a quality determination unit that determines quality based on a difference between a molecule specified in the reference data and a molecule specified in the target data.

4. A morphological analysis device according to any one of claims 1 to 3, characterized in that the parameters representing the morphology represent physical differences that directly result from chemical differences in the components of the molecule identified in the target data compared to the molecule identified in the reference data.

5. A morphological analysis device according to any one of claims 1 to 3, characterized in that the parameters representing the morphology represent differences resulting from at least one of the conformation and mobility of the molecule identified in the target data compared to the molecule identified in the reference data.

6. 6. The morphological analysis apparatus according to claim 1, wherein the molecules identified in the subject data are identical to the molecules identified in the reference data at different times.

7. 7. The morphological analysis device according to claim 1, wherein the morphology estimation unit identifies the distribution of physical quantities constituting the molecular shape by the level of the physical quantities constituting the molecular shape, or the shape or size of the spread expressed as an aggregate of these quantities.

8. 8. The morphology analysis device according to claim 1, wherein the morphology estimation unit uses, as the reference data, data on physical quantities constituting molecular shapes obtained by measuring a standard sample with low mobility.

9. 4. The morphological analysis device according to claim 3, wherein the morphological estimation unit or the quality determination unit has an AI function.

10. an output information generating unit that generates output information so that the form of the target data can be recognized; the output information generation unit displays the target data on an output device, and receives input of designation information that designates a component whose form is to be estimated for the displayed target data; 10. The morphological analysis device according to claim 1, wherein the morphological estimation unit compares parameters indicated by the morphology of the component specified based on the specification information.

11. an output information generating unit that generates output information so that the form of the target data can be recognized; 10. The morphological analysis device according to claim 1, wherein the output information generation unit displays the target data on an output device and further displays a specific mark indicating the estimated morphology superimposed on the target data.

12. 10. The morphological analysis apparatus according to claim 1, wherein the test molecule is a Y-shaped antibody molecule, and the constituent element is Fab or Fc.

13. an output information generating unit that generates output information so that the form of the target data can be recognized; a quality determination unit that determines whether the test molecule meets a quality standard based on the estimated morphology, 13. The morphological analysis apparatus according to claim 12, wherein the output information generating unit displays, when the test molecule does not conform to the quality standard, that the test molecule does not conform to the quality standard.

14. A morphology analysis method for analyzing the morphology of a molecule that is placed in a state where it can exist in multiple molecular shapes, comprising: reading out reference data, which is molecular shape data of one of the test molecules; a step of reading out target data, which is molecular shape data of the other test molecule; a step of identifying parameters indicating the shape of the molecular shapes of the reference data and the target data from a distribution of physical quantities, such as an electron density or potential map, obtained at a resolution of 3 Å or more and 10 Å or less, which constitute the molecular shapes of the reference data and the target data, and comparing the parameters indicating the shape identified from the reference data with the parameters indicating the shape identified from the target data; and determining whether or not a molecule identified in the reference data matches a molecule identified in the target data based on a result of the comparison; A morphological analysis method characterized in that both the reference data and the target data have molecular level resolution.

15. A morphology analysis program for analyzing the morphology of a molecule that is placed in a state where it can exist in multiple molecular shapes, On the other hand, a process of reading out reference data, which is molecular shape data of the test molecule, On the other hand, a process of reading out target data, which is molecular shape data of the test molecule; a process of identifying parameters indicating the shape of each of the reference data and the target data from a distribution of physical quantities, such as an electron density or potential map, obtained at a resolution of 3 Å or more and 10 Å or less, which constitute the molecular shapes of the reference data and the target data, on the assumption that the reference data and the target data are data on a plurality of possible molecular shapes obtained as an ensemble, and comparing the parameters indicating the shape identified from the reference data with the parameters indicating the shape identified from the target data; the morphological analysis device causes a computer to execute a process of determining whether or not a molecule identified in the reference data matches a molecule identified in the target data based on a result of the comparison; A morphological analysis program characterized in that both the reference data and the target data have molecular level resolution.

Images