Object abnormality determination system and object abnormality determination method

Abstract

To provide an abnormality determination system of an object capable of determining the presence or absence of abnormality of the object without using a high-level device or performing complicated processing.SOLUTION: An object abnormality determination system for solving the above-described problem includes a signal generation unit for generating a plurality of signals based on an interaction between an object and two or more types of luminescent probes, a detection unit for acquiring the plurality of signals, and an analysis unit for analyzing the plurality of signals acquired by the detection unit and determining presence or absence of an abnormality of the object. Each of the light-emitting probes has a light-emitting behavior changed by interaction with the object. The analysis unit determines the presence or absence of an abnormality of the target object with reference to a learned model obtained by performing machine learning on only a plurality of signals based on an interaction between a reference substance corresponding to the normal target object and the two or more types of luminescent probes.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a system for determining an abnormality in an object and a method for determining an abnormality in an object. [Background technology]

[0002] Anomaly detection and screening are extremely important in the quality control and distribution management of various products. For example, in the manufacturing process of a product, anomalies can occur in the product due to various factors such as the manufacturing environment, raw material lot, storage conditions, etc. To detect such anomalies, various inspections are carried out depending on the assumed cause and defect.

[0003] Meanwhile, in recent years, the distribution of counterfeit medicines has become a problem, with highly sophisticated counterfeit medicines circulating on the market. Various inspection methods have been proposed to identify these counterfeit medicines. For example, Non-Patent Document 1 proposes analyzing impurity profiles using high performance liquid chromatography (HPLC). Also, Non-Patent Document 2 proposes identifying counterfeit medicines using Raman scattering analysis, etc. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Tanimoto, "Research on Identification Methods for Counterfeit Medicines - Determining the Authenticity of Sildenafil Citrate Preparations by Impurity Profiles," Ministry of Health, Labour and Welfare Sciences Research Grant-Supported Research Report, March 15, 2018 [Non-patent document 2] Yoshida, "Research into improving the accuracy of counterfeit drug identification methods using impurity profile analysis," Grant-in-Aid for Scientific Research, Research Results Report, May 24, 2016 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the method described in Non-Patent Document 1 has the problem that it is difficult to determine the authenticity of a sample with an unknown impurity profile. Furthermore, when a generic product (genuine product) exists, there are few cases where the impurity profiles of the original product (genuine product) and the generic product completely match. Therefore, even if the impurity profile of a sample is analyzed, there is also the problem that it is not possible to distinguish whether the sample is a generic product or a counterfeit product.

[0006] Furthermore, the Raman scattering analysis method described in Non-Patent Document 2 is an advanced technology, and the cost of the equipment is high. Furthermore, specialized knowledge is required to operate the equipment and analyze the data. Furthermore, this method makes it difficult to distinguish between genuine products and counterfeit products, which contain very similar ingredients. It is conceivable that manufacturers of counterfeit products may exploit this characteristic by adding similar ingredients to pass inspections based on Raman scattering analysis.

[0007] The present invention has been made in view of the above-mentioned problems. That is, an object of the present invention is to provide an abnormality detection system and an abnormality detection method for an object that can determine whether or not an object has an abnormality without using advanced devices or performing complex processing. [Means for solving the problem]

[0008] In order to achieve at least one of the above-mentioned objects, the following system and method for determining an abnormality in an object are provided.

[0009] One embodiment of the present invention is an abnormality discrimination system for discriminating whether or not an object is abnormal, the system including: a signal generation unit for generating a plurality of signals based on interactions between the object and two or more types of luminescent probes; a detection unit for acquiring the plurality of signals; and an analysis unit for analyzing the plurality of signals acquired by the detection unit and discriminating whether or not an abnormality is present in the object, wherein the luminescent behavior of each of the luminescent probes changes due to interaction with the object, and the analysis unit discriminates whether or not an abnormality is present in the object by referring to a trained model that has been machine-learned using only a plurality of reference signals based on interactions between a reference substance corresponding to a normal object and the two or more types of luminescent probes.

[0010] One embodiment of the present invention further provides a method for determining whether or not an object has an abnormality, the method comprising the steps of: allowing the object to interact with two or more types of luminescent probes; generating a plurality of signals from the two or more types of luminescent probes that have interacted with the object; acquiring the plurality of signals; and analyzing the acquired plurality of signals to determine whether or not an abnormality exists in the object, wherein the luminescent behavior of each of the luminescent probes changes upon interaction with the object; and the step of determining whether or not an abnormality exists in the object further provides a method for determining whether or not an abnormality exists in the object by referring to a trained model that has been machine-learned using only a plurality of reference signals based on interactions between a reference substance corresponding to a normal object and the two or more types of luminescent probes. [Effects of the Invention]

[0011] According to the system and method for determining whether an object is abnormal, which relate to one embodiment of the present invention, it is possible to easily determine whether an object has an abnormality or not, without using advanced equipment or performing complex processing. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a flowchart of a method for determining abnormality in an object according to one embodiment of the present invention. [Figure 2]FIG. 2 is a schematic diagram of a system for determining abnormality in an object according to one embodiment of the present invention. [Figure 3] 3A and 3B show the results of the restoration error when the method for determining abnormality in an object is performed in the example. [Figure 4] FIG. 4 is a graph showing the contribution of the principal components when the method for determining abnormality in an object is performed in the example. DETAILED DESCRIPTION OF THE INVENTION

[0013] The present invention will be described in detail below based on embodiments, but the present invention is not limited to these embodiments.

[0014] The object anomaly detection system and object anomaly detection method of the present embodiment determine whether an object has an anomaly. In this specification, "an anomaly in an object" refers to an object that does not meet a desired standard, does not meet a desired quality, or is not a genuine product. As described above, various methods have been used to detect anomalies in various products, intermediates, etc., but these methods often require complex processing and equipment. Furthermore, when it is unclear what anomaly has occurred in an object or how a counterfeit product differs from a genuine product, it is difficult to select an inspection method and to properly identify anomalies. Furthermore, when counterfeit product manufacturers take measures to evade inspection, accurate anomaly detection is difficult.

[0015] In contrast, in the system and method for detecting an abnormality in an object according to the present embodiment, two or more types of luminescent probes are caused to interact with each other in the object, generating multiple signals. The multiple signals are then analyzed to detect an abnormality in the object. This method (system) does not detect an abnormality by focusing only on specific physical properties of the object, but rather performs the detection based on information that cannot be interpreted from the signal alone. Therefore, it is characterized in that it is difficult for manufacturers of counterfeit goods to take countermeasures. Furthermore, because the analysis is performed based on information that is not specialized in specific physical properties and that cannot be interpreted from the signal alone, it is possible to identify an abnormality even when an unexpected abnormality occurs in the object.

[0016] Furthermore, in this embodiment, when determining whether an object is abnormal, a trained model is referenced, which is machine-learned from only multiple reference signals based on interactions between a reference substance corresponding to a normal object and two or more types of luminescent probes, i.e., only ground truth data. This method (system) can accurately acquire information about an object (reference substance) in a normal state, allowing for highly accurate determination of whether an object is abnormal. Furthermore, if an unknown change occurs in the object, it is also possible to determine whether the change is within an acceptable range.

[0017] Furthermore, according to this embodiment, a specific physical property of the object is not measured, which has the advantage that there is no need to select a signal type or a measurement means to be used for anomaly detection for each object.

[0018] Here, the type of object for which an abnormality can be detected by the object abnormality detection system and object abnormality detection method of this embodiment is not particularly limited, as long as it is capable of interacting with the luminescent probe. For example, it may be a chemical substance such as a pharmaceutical, or agricultural products or processed products thereof. It may also be biological substances or their metabolites, or even soil, river water, seawater, etc. The object may be in any state, such as gas, liquid, or solid. However, it is preferable to dissolve gas or solid in an appropriate solvent before detecting an abnormality, in order to ensure sufficient interaction with the luminescent probe. Below, the object abnormality detection method will be described first, followed by the object abnormality detection system.

[0019] 1. Method for determining abnormalities in an object The flow of the method for discriminating an abnormality in an object according to this embodiment is shown in Figure 1. This method for discriminating an abnormality in an object includes the steps of: a step (interaction step) S101 of causing an interaction between the object and two or more types of luminescent probes; a step (signal generation step) S102 of generating a plurality of signals from the two or more types of luminescent probes that have interacted with the object; a step (signal acquisition step) S103 of acquiring the plurality of signals; and a step (discrimination step) S104 of analyzing the acquired plurality of signals with reference to a trained model created in advance and discriminating whether or not an abnormality exists in the object. Note that the method for discriminating an abnormality in an object according to this embodiment may further include steps other than those described above, provided that the purpose and effects of this embodiment are not impaired.

[0020] (Interaction step S101) In the interaction step S101, the target substance is allowed to interact with two or more types of luminescent probes. Here, each luminescent probe may be a compound whose luminescent behavior changes upon interaction with the target substance. The luminescent probe may be, for example, a compound having a binding moiety that chemically and / or physically binds to the target substance and a luminescent moiety whose luminescent behavior changes depending on the structure and / or state of the target substance. The binding moiety of the luminescent probe may specifically bind to a specific structure of the target substance, or may bind non-specifically to the target substance. Specific examples of luminescent probes will be described in detail later.

[0021] The number of luminescent probes that are allowed to interact with the target substance needs only to be two or more types, and is selected appropriately depending on the type of target substance to be discriminated, etc. If 50 or more types of luminescent probes are used, more preferably 100 or more types, a very detailed analysis can be performed in the discrimination step S104 described below, further increasing the accuracy of abnormality discrimination.

[0022] In this embodiment, two or more types of luminescent probes may be allowed to interact with the target collectively, but it is preferable to allow multiple luminescent probes to interact with the target individually. When the luminescent probes are allowed to interact individually, it becomes possible to individually acquire signals from each luminescent probe in the signal acquisition step described below, increasing the amount of data. Therefore, the accuracy of abnormality discrimination in the discrimination step is further improved.

[0023] The method for causing an interaction between the target and the luminescent probe is not particularly limited, and the target and the luminescent probe may simply be mixed. Furthermore, a solvent, a dispersion medium, or other medium may be used as needed. A specific example of a method for causing an interaction between the target and the luminescent probe is a method in which a plurality of luminescent probes are placed individually or collectively in the wells of a microwell plate, and a liquid containing the target is injected into the well to cause them to interact. Alternatively, the target may be placed first, and then a liquid containing the luminescent probe is injected later. Another example is a method in which luminescent probes are immobilized on a microplate or a microarray, and the target is introduced onto the microplate or microarray by a conventional method. In this case, the target may also be immobilized first, and then the luminescent probe is introduced later.

[0024] (Signal generation step S102) In the signal generation step S102, multiple signals are generated from two or more types of luminescent probes that have interacted with the target. There are no particular limitations on the type of signal, as long as it is useful for analysis in the discrimination step described below. As described above, the luminescent behavior of a luminescent probe changes when it interacts with the target. Therefore, it is preferable to irradiate each luminescent probe that has interacted with the target with excitation light, causing each luminescent probe to generate light (a signal). In this case, the wavelength of the irradiated excitation light and the irradiation method of the excitation light are appropriately selected depending on the signal acquisition method in the signal acquisition step, the type of signal used in the analysis, etc. Furthermore, light of a single wavelength may be irradiated only once or multiple times as excitation light. Furthermore, light of different wavelengths may be irradiated as excitation light, either all at once or in separate multiple times.

[0025] (Signal acquisition step S103) In the signal acquisition step S103, two or more types of signals emitted from the two or more types of luminescent probes are acquired. The method of acquiring the signals is not particularly limited and is selected appropriately depending on the type of signal. For example, when acquiring light emitted by each luminescent probe, the luminance or wavelength may be acquired as a signal. Furthermore, changes over time in the spectral distribution or changes over time in chromaticity of the light emitted by the luminescent probe may also be acquired.

[0026] (Discrimination step S104) In the discrimination step S104, the plurality of signals (discrimination data) acquired in the signal acquisition step S103 for the object are analyzed with reference to a trained model prepared in advance, and the presence or absence of an abnormality in the object is discriminated.

[0027] The creation (step) of the trained model used in the discrimination step S104 is performed before the analysis step. Specifically, the trained model is created using the following procedure. First, a reference substance corresponding to a normal object is allowed to interact with two or more types of luminescent probes. The luminescent probes used here are preferably the same as those used in the interaction step S101. The reference substance may also be anything that corresponds to a normal object. For example, if the purpose of discriminating between abnormalities in an object is to discriminate between authenticity and counterfeit medicines, an authentic product is used as the reference substance. If the purpose of discriminating between abnormalities in a target substance is quality control, a product that meets the desired quality standards is used as the reference substance.

[0028] The number of reference materials used to create a trained model may be only one, but two or more are preferable, and the more the better. For example, when the target is a chemical substance (such as a pharmaceutical), acceptable variations generally occur during the manufacturing and distribution processes. Therefore, by acquiring data on many reference materials and performing machine learning on this data, the misclassification rate in the discrimination step S104 can be reduced. Furthermore, by using a large number of reference materials with various conditions (e.g., manufacturing lot, storage conditions, raw material lot), the misclassification rate can be further reduced. Similarly, when the target is an agricultural product, a biological material, or a material derived from these, it is preferable to use a large number of reference materials with various conditions. Here, the number of reference materials used to create a trained model is appropriately selected depending on the type of target and the purpose of abnormality discrimination. Typically, 10 or more reference materials are preferable from the perspective of appropriately learning the distribution of normal reference materials and acceptable variations. Furthermore, 30 or more reference materials are more preferable from the perspective of stabilizing the distribution and more accurately learning the characteristics of the population. From the viewpoint of reducing the rate of misclassification and increasing the accuracy of abnormality discrimination in the discrimination step S104, and from the viewpoint of making it easier to define the boundary between abnormality and normality more precisely, 100 or more is more preferable.

[0029] Next, a plurality of reference signals are generated from two or more types of luminescent probes that have been allowed to interact with the reference substance, and the plurality of reference signals are acquired. The methods for generating and acquiring the reference signals are the same as those described in the signal generation step S102 and signal acquisition step S103 above. Then, a learning model is constructed by machine learning the large number of reference signals that have been obtained. In this case, the machine learning method is not particularly limited, and examples that can be used include statistical methods such as principal component analysis, boxplots, and statistical tests, as well as methods such as clustering, isolated forests, one-class support vector machines (SVM), autoencoders, and deep learning.

[0030] In the discrimination step S104, the trained model obtained as described above is referenced and the signal of the object acquired in the signal acquisition step S103 is analyzed. For example, the trained model and the signal of the object acquired in the signal acquisition step S103 are compared and calculated, and if the signal significantly deviates from the characteristics of the trained model, it is determined that there is an abnormality. Similarly, if the signal has the characteristics of the trained model or is close to them, it is determined that there is no abnormality.

[0031] The threshold for identifying anomalies can be set appropriately depending on the type of object and the purpose of the identification. For example, if you want to reduce the probability of false negatives, you can set it so that only cases that fit the characteristics of the trained model are identified as normal. In particular, when there is a very high risk of abnormal products being mixed in (for example, in pharmaceutical or food inspections), you can set the threshold so that an object is identified as abnormal even when the object data corresponds to data at the extreme positions in the distribution of the trained model. This method may result in an object being identified as abnormal even when it is normal. However, it can reliably reduce false negatives. On the other hand, if you want to reduce the probability of false positives, you can set it so that an object is identified as normal up to a range slightly away from the characteristics of the trained model.

[0032] (About luminescent probes) The luminescent probe used in the analysis method of this embodiment can be a compound having a binding moiety in its molecule that chemically and / or physically binds to at least a portion of the target substance, and a luminescent moiety whose luminescence behavior changes depending on the structure and / or state of the target substance. The luminescent probe may have only the binding moiety and the luminescent moiety. Alternatively, the luminescent probe may further have a structure that does not contribute to binding to the target substance or luminescence. However, the binding moiety and the luminescent moiety are preferably located on the tip side of the luminescent probe, i.e., the side that is most likely to come into contact with the target substance.

[0033] Here, the binding moiety is selected depending on the type of target substance, but it is preferable that the luminescent probe contains a nucleic acid structure. When the luminescent probe contains a nucleic acid structure, the luminescent probe can be physically and / or chemically bound to proteins, or peptides and amino acids that constitute proteins. Note that, in this specification, the term "nucleic acid structure" includes not only structures derived from DNA or RNA, but also structures derived from phosphorothioate oligodeoxynucleotides, 2'-O-(2-methoxy)ethyl-modified nucleic acids, siRNA, cross-linked nucleic acids, peptide nucleic acids, aTNA, SNA, GNA, LNA, and morpholino antisense nucleic acids.

[0034] Furthermore, the type of emission of the light-emitting moiety is not particularly limited as long as its emission behavior changes depending on the interaction with the target, for example, the structure or state of the target. The light-emitting moiety may emit only one type of light in response to a single excitation light. However, it is preferable that the light-emitting moiety exhibit two or more types of emission selected from the group consisting of fluorescence, phosphorescence, excimer emission, exciplex emission, thermally activated delayed fluorescence, excited-state intramolecular proton emission, triplet triplet annihilation emission, twisted intramolecular charge transfer emission, and aggregation-induced emission in response to a single excitation light. When the light-emitting moiety exhibits such two or more types of emission, more information can be acquired in the signal acquisition step described above. For example, when a luminescent probe exhibiting such two or more types of emission is bound to the target, the structure and electronic state of the luminescent moiety change, resulting in complex emission behavior different from that of the luminescent probe alone. For example, when a luminescent probe exhibiting three different types of emission (fluorescence, phosphorescence, and excimer emission) in response to a single excitation light is allowed to interact with the target, the two or more types of emission occur. The processes that produce fluorescence, phosphorescence, and excimer emission change, and the wavelength and lifetime of each light change. Therefore, depending on the structure of the object, a complex and large amount of data can be obtained by combining these lights. This complex and large amount of data makes it possible to understand the structure and state of the object in great detail.

[0035] Specific examples of luminescent probes include those having a binding moiety consisting of a nucleic acid structure and at least two chromophores or luminophores (luminescent moieties) bound to the backbone of the nucleic acid structure. More specifically, these include molecules having a backbone with one or more structural units containing a pentose- or hexose-derived sugar structure and a phosphate ester bond bound to the sugar structure, and one or more chromophores or luminophores bound to the sugar structure.

[0036] The main chain of the binding portion of the luminescent probe may have one or more structural units containing a sugar structure derived from a pentose or hexose and a phosphate ester bond bound to the sugar structure. The main chain may contain only one of the structural units, or may contain multiple structural units. That is, the main chain may have one of the sugar structures and one of the phosphate ester bonds bound to the sugar structure, or may have a structure containing the sugar structure and phosphate ester bonds alternately. Usually, both ends of the main chain of the luminescent probe are sugar structures, so that the number of sugar structures is one more than the number of phosphate ester bonds. When the main chain contains multiple structural units, the multiple structural units may be the same or different from each other.

[0037] The number of the structural units contained in the main chain of the binding moiety is selected appropriately depending on the type of target substance, etc., but is preferably 2 or more and 6 or less. As the amount of the structural units increases, the binding moiety becomes more likely to specifically bind to the target substance. However, it is preferable that the luminescent probe has appropriate (not excessive) specificity for the target substance, and the number of the structural units is preferably 6 or less. By using multiple probes that are not excessively specific or have excessive binding strength, it becomes easier to detect various structural changes compared to probes that react only to a specific three-dimensional structure.

[0038] The main chain of the bond may partially contain a structure other than the structural unit containing the pentose- or hexose-derived sugar structure and the phosphate ester bond, as long as the purpose and effect of this embodiment are not impaired. The structures at both ends of the main chain are not particularly limited and may be various structures, such as an OH group or an alkoxy group.

[0039] Examples of the pentose include ribose, deoxyribose, and xylose. Specific examples of the hexose include allose, glucose, and mannose. Among these, a sugar structure derived from ribose or deoxyribose is particularly preferred, since the main chain of the luminescent probe has a structure similar to that of DNA or RNA, which facilitates interaction with the target substance.

[0040] When the structural unit contains a structure derived from ribose or deoxyribose, the phosphate ester bond is preferably bonded to the 3rd and 5th carbons of the ribose or deoxyribose. The luminescent moiety described below is preferably bonded to the 1st position of the ribose or deoxyribose. That is, the luminescent probe preferably contains a structure represented by the following general formula (1a) or (1b): [ka] In the general formulae (1a) and (1b), Y represents a light-emitting moiety described below.

[0041] On the other hand, the main chain of the binding portion is not limited to a structural unit containing a sugar structure derived from pentose or hexose and a phosphate ester bond. Representative examples of other structural units include peptide nucleic acid structural units such as those shown below. [ka] Peptide nucleic acids have no charge and are free from electrostatic repulsion, allowing them to form stronger associations with target molecules. Furthermore, they are resistant to enzymes such as nucleases and proteases, making them suitable for use in cells. Furthermore, they can be synthesized on a relatively large scale.

[0042] On the other hand, the light-emitting portion (chromophore or luminophore) may have a structure that emits a predetermined type of light by itself in response to a single excitation light, or that emits a predetermined light by the action of multiple chromophores or luminophores. In this specification, the term "chromophore" refers to a structure that absorbs light with a wavelength of 300 nm or more, and the term "luminophore" refers to a structure that absorbs light with a wavelength of 300 nm or more and emits light.

[0043] The number of chromophores or luminophores possessed by each luminescent probe may be only one, as long as the luminescent probe is capable of exhibiting multiple types of luminescence. However, from the viewpoint that the luminescent probe is likely to exhibit multiple types of luminescence, two or more are preferred, and three to six are even more preferred. When a luminescent probe has multiple chromophores or luminophores, the types may be only one, or two or more. Usually, one chromophore or luminophore is bound to one sugar structure at the binding moiety. Therefore, when a luminescent probe has two or more chromophores or two or more luminophores, it is preferable that the sugar structures in the main chain of the binding moiety also be two or more.

[0044] In addition, when the number of chromophores or luminophores in the luminescent probe is less than the number of sugar structures (or peptide structures) in the binding portion, some sugar structures will have no chromophores or luminophores bound to them. The sugar structures to which no chromophores or luminophores are bound may not have other atomic groups bound to them, and may have natural nucleic acid bases bound to them. In this specification, natural nucleic acid bases refer to adenine, guanine, cytosine, thymine, and uracil. However, the total number of natural nucleic acid bases bound to the main chain is preferably 50% or less, more preferably 25% or less, of the total number of structural units constituting the binding portion of the luminescent probe. When the number of natural nucleic acid bases is 50% or less, association between luminescent probes is suppressed, and binding between the target and the luminescent probe is more likely to be dominant.

[0045] Here, examples of luminophores that emit fluorescence include structures derived from fluorescein, rhodamine, boron dipyrromethene, etc. Examples of luminophores that emit phosphorescence include structures derived from iridium complexes, platinum complexes, etc. Examples of luminophores that emit excimer emission include structures derived from pyrene, anthracene, perylene, etc. Examples of luminophores that emit exciplex emission include structures derived from pyrene-dimethylaniline, etc. Examples of luminophores that emit thermally activated delayed fluorescence include structures derived from 4CzIPN, DABNA, etc. Examples of luminophores that emit excited-state intramolecular proton emission include structures derived from hydroxyphenylbenzoxazole, etc. Examples of luminophores that emit triplet triplet annihilation emission include structures derived from 9,10-diphenylanthracene, rubrene, etc. Examples of luminophores that emit twisted intramolecular charge transfer emission include structures derived from diaminoanthracene, diaminonaphthalene, etc. Examples of lumophores that emit aggregated organic luminescence include structures derived from tetraphenylethene, hexaphenylsilole, and the like.

[0046] Furthermore, luminescent compounds used as light-emitting materials or hosts, electron transport materials, hole transport materials, or light-emitting materials in organic electroluminescence (EL) can also be suitably used as materials for the chromophores or luminophores. Specific examples of such luminescent compounds include those described in "State-of-the-art Organic EL" (CMC Publishing), "Organic EL Material Technology" (CMC Publishing), "All About Organic EL" (Nippon Jitsugyo Publishing), and "Various Dye Materials Pioneering the Future" (Kagaku Dojinsha). Furthermore, the luminescent probe may further include a structure that exerts various functions as a site for controlling the interaction between the binding moiety and the target substance.

[0047] In this embodiment, the luminescent probe preferably contains, as a luminophore, at least one structure selected from a structure that emits fluorescence, a structure that emits excimer luminescence, and a structure that emits exciplex luminescence. It is particularly preferable that the luminescent probe contains at least a structure that emits fluorescence. When the luminescent probe emits fluorescence, this has the advantage of being easy to analyze using various measuring devices.

[0048] Furthermore, the luminescent probe preferably exhibits multiple types of luminescence when irradiated with light having a wavelength of 300 to 400 nm. If the luminescent probe exhibits multiple types of luminescence when irradiated with light of the wavelength, a special light source is not required when analyzing the target object, and the target object is less likely to be damaged.

[0049] However, when an LED or organic EL element is used as the excitation light source, excitation in the visible light range is advantageous. Therefore, when using such a light source, the absorption wavelength of the luminescent probe is preferably 400 to 700 nm.

[0050] The molecular weight of the luminescent probe is selected appropriately depending on the type of binding moiety and luminescent moiety possessed by the luminescent probe, and is generally preferably from 500 to 10,000, and more preferably from 500 to 4,000. When the molecular weight of the luminescent probe is 10,000 or less, the specificity for the target substance is appropriately low, making it possible to cause the luminescent probe to react nonspecifically with multiple sites on the target substance.

[0051] The method for producing the luminescent probe is selected appropriately depending on the structure of the binding site in the luminescent probe. For example, a luminescent probe having the above-mentioned sugar structure can be produced by the following method. A monomer is prepared in which the above-mentioned chromophore or luminophore and a phosphate ester are bound to a pentose or hexose. The monomer can be synthesized by polymerizing the desired sequence using the phosphoramidite method with a DNA / RNA synthesizer or the like. This method allows multiple types of monomers with different types of chromophores or luminophores to be prepared, and the desired number of monomers can be bound by changing the sequence order. In other words, a wide variety of luminescent probes can be synthesized from multiple types of monomers with different types of chromophores or luminophores. By changing the type of monomer used and the number of monomers bound, a very large number of luminescent probes can be synthesized.

[0052] 2. System for detecting abnormalities in objects FIG. 2 shows a schematic diagram illustrating the configuration of an object anomaly detection system for performing the above-described object anomaly detection method. However, the configuration of the anomaly detection system is not limited to this configuration. The analysis system 100 shown in FIG. 2 includes a signal generation unit 11 for generating multiple signals based on the interaction between the object and two or more types of luminescent probes, a detection unit 12 for acquiring the multiple signals from the signal generation unit 11, and an analysis unit 13 for analyzing the multiple signals acquired by the detection unit 12 with reference to the trained model described above and determining whether or not an abnormality exists in the object. The analysis system 100 may include other components depending on its application. Each component will be described below.

[0053] (Signal generation section) The signal generation unit 11 is configured to generate a signal based on the interaction between the target object and two or more types of luminescent probes. The structure of the signal generation unit 11 is selected appropriately depending on the type of signal to be generated. The signal generation unit 11 of this embodiment irradiates light onto the luminescent probes, causing them to emit light. The signal generation unit 11 has a light source 111, a storage unit 112 for storing the luminescent probes and the target object, and an optical system 114 for guiding the light from the light source 111 to the storage unit 112 (or to the luminescent probes stored in it).

[0054] The light source 111 is not particularly limited as long as it is a means capable of irradiating the luminescent probe that has interacted with the target object with light of a desired wavelength for a desired period of time. Examples of preferred light sources include picosecond diode lasers, tunable lasers, supercontinuum light sources, and LED light sources. These light sources 111 can irradiate the luminescent probe with light of a predetermined wavelength for only a short period of time. In consideration of the signal-to-noise ratio (SN) in the detection unit 12, it is preferable to select a light source that can be quenched before the luminescent probe emits light.

[0055] The storage section 112 is not particularly limited as long as it has a structure capable of storing the above-mentioned luminescent probes and target objects. Examples of the storage section 112 include a microwell plate, a microplate, a microarray, and the like. Note that the luminescent probes may be stored in advance in these storage sections 112. Alternatively, the target objects may be stored in the storage sections 112.

[0056] The optical system 114 is not particularly limited as long as it is capable of guiding light from the light source 111 to the housing section 112 that houses the luminescent probe and the target object, and is capable of guiding light emitted by the luminescent probe to the detection section 12. The optical system 114 may have, for example, an excitation light filter for cutting light of unnecessary wavelengths emitted from the light source 111. The optical system 114 may also have a dichroic mirror that reflects light from the light source 111 to the housing section 112, while transmitting light emitted by the luminescent probe. The optical system 114 may also have an optical filter or the like that cuts light of unnecessary wavelengths from the light that has passed through the dichroic mirror.

[0057] (Detection unit) The detection unit 12 is not particularly limited as long as it is a means capable of acquiring each of the multiple signals (here, multiple lights) emitted by the multiple luminescent probes. It is appropriately selected depending on the type of signal to be acquired. When the signal is light, as in this embodiment, it may be a known camera or the like. It may also be, for example, a CCD camera, CMOS camera, or the like that captures images intermittently or continuously.

[0058] (Analysis Department) The analysis unit 13 may be any means capable of analyzing the multiple signals acquired by the above-described detection unit 12 with reference to a trained model. The analysis unit 13 may be, for example, a means for reading out the trained model from an external storage device (not shown) or an internal storage means (not shown), and performing a comparative calculation of the data (analysis data) acquired by the above-described detection unit with the trained model.

[0059] A general-purpose computer equipped with storage means such as a hard disk drive (HDD), solid state drive (SSD), or read-only memory (ROM) for storing programs, data, etc., and a central processing unit (CPU) for executing programs, performing calculations, etc., can be used as the analysis unit 13. The computer may further include input means such as a keyboard and a mouse, and output means such as a monitor and a printer. [Example]

[0060] 1. Luminescent Probe Preparation Eighty-one types of luminescent probes were prepared, as shown in the chemical formula below. Each of the 81 types of luminescent probes has a structure in which 13 types of groups (luminophores, chromophores, or other groups) shown on the right side are bonded to the R position of the structure on the left side of the chemical formula below. The 81 types are composed of combinations of groups represented by R. [ka]

[0061] 2. Preparation of training data and detection of abnormalities in the target object (1) Placement of luminescent probes Seventy-five 96-well microwell plates (hereinafter also referred to as "well plates") were prepared, each with wells having an opening diameter of 7 mm arranged in 12 columns and 8 rows at 9 mm intervals. Eighty-one types of luminescent probes were sequentially placed in the wells of each well plate. The 81 types of luminescent probes were arranged so that only one luminescent probe was placed in each well, and the same luminescent probe was placed in the corresponding position of each well plate.

[0062] (2) Placement of the reference material and the target object (interaction step) The samples shown in Table 1 below were prepared as reference materials (for training data) and targets (for test data) for creating a trained model. Specifically, commercially available pharmaceuticals (tablets) and counterfeit products made to resemble them were each crushed into powder and dissolved in dimethyl sulfoxide (DMSO) to the concentrations shown in Table 1 (samples 1 to 24). For comparison, a sample containing only DMSO was also prepared (sample 25). To ensure accuracy, three samples of each type were prepared. Samples A1, A2, A3, B1, B2, C1, and C2 in Table 1 are as follows: Sample A1 (genuine over-the-counter medicine) Sample A2 (Sample A1 left in a 40°C / 80% RH environment for 24 hours) Sample A3 (same drug as Sample A1, but different acquisition date and lot number) Sample B1 (a counterfeit product claiming to have the same effect as Sample A1) Sample B2 (the same counterfeit product as Sample B1, but acquired at a different time and lot) Sample C1 (a counterfeit product that claims to have the same effect as Sample A1, but is different from Sample B1) Sample C2 (the same counterfeit product as Sample C1, but acquired at a different time and lot) [Table 1]

[0063] The 75 samples (25 types x 3) were placed in each well of 75 well plates containing luminescent probes, with a different sample placed on each well plate.

[0064] (3) Signal generation process and signal acquisition process The microwell plate was irradiated with excitation light (wavelength 365 nm) (signal generation step). The fluorescence spectrum at this time was photographed with a camera, and RGB information for each microwell plate was obtained (signal acquisition step). The same procedure was performed for all microwell plates.

[0065] (4) Trained model creation process Of the data acquired in the signal acquisition step, data for samples 1 to 6, 8 to 13, and 15 to 20 were used as data for creating a trained model. Principal component analysis was performed on these data, and machine learning was performed up to an arbitrary dimension.

[0066] (5) Process for detecting abnormalities in the object Of the data acquired in the signal acquisition process, data on sample 7 (genuine), sample 14 (genuine), samples 21-24 (counterfeits), and sample 25 (for comparison) were subjected to anomaly discrimination. Specifically, the restoration error when a matrix transformation similar to that for the genuine product was performed was used as an index for comparing the degree of similarity with the characteristics of the genuine product, and any deviation from the characteristics of the genuine product was classified as anomaly (counterfeit). The results are shown in Figures 3A and 3B. Note that Figure 3B is a graph that enlarges a portion of Figure 3A (the mountain on the left).

[0067] As shown in Figures 3A and 3B, the authentic products (samples 7 and 14) showed a very high degree of similarity to the training data. On the other hand, the counterfeit products (samples 21 to 24) showed a large difference from the training data. In other words, it was possible to distinguish between authentic and counterfeit products.

[0068] Figure 4 shows the contribution rate of the principal components of the training data in the anomaly detection process. As shown in Figure 4, anomaly detection of an object is not performed by only a specific principal component, but about 90% of the information is explained by five principal components. In other words, it can be said that anomaly detection is performed by multifaceted information. [Industrial Applicability]

[0069] The above-described analysis system and analysis method make it possible to detect abnormalities in an object without using special equipment or complex preprocessing. Therefore, they are useful for quality control and production management in various manufacturing fields. Furthermore, they are also very useful for determining the authenticity of over-the-counter drugs. [Explanation of symbols]

[0070] 11 Signal generation unit 12 Detector 13 Analysis Department 111 Light source 112 Storage unit 114 Optical system 100 Analysis System

Claims

1. An abnormality detection system for determining whether or not an object has an abnormality, a signal generating unit for generating a plurality of signals based on the interaction between the target substance and two or more types of luminescent probes; a detector for acquiring the plurality of signals; an analysis unit that analyzes the plurality of signals acquired by the detection unit and determines whether or not there is an abnormality in the object; Including, the luminescent behavior of each of the luminescent probes changes upon interaction with the target substance; the analysis unit refers to a trained model that has been machine-learned using only a plurality of reference signals based on interactions between a reference substance corresponding to a normal object and the two or more types of luminescent probes, and determines whether or not the object is abnormal. A system for detecting abnormalities in objects.

2. The luminescent probe comprises a binding portion that chemically and / or physically binds to at least a part of the target, and a luminescent portion whose luminescence behavior changes depending on the structure and / or state of the target. The system for determining an abnormality in an object according to claim 1 ,

3. the trained model is a model obtained by machine learning of a plurality of reference signals based on interactions between a plurality of the reference substances and the two or more types of luminescent probes; The system for determining abnormality in an object according to claim 1 .

4. The object is a pharmaceutical product. The system for determining abnormality in an object according to claim 1 .

5. It is a method for determining whether an object has an abnormality or not, A step of allowing the target substance to interact with two or more types of luminescent probes; generating a plurality of signals from the two or more types of luminescent probes that have interacted with the target substance; acquiring the plurality of signals; analyzing the acquired signals and determining whether or not there is an abnormality in the object; Including, the luminescent behavior of each of the luminescent probes changes upon interaction with the target substance; In the step of determining whether or not the object has an abnormality, the presence or absence of the abnormality of the object is determined by referring to a trained model that has been machine-learned using only a plurality of reference signals based on interactions between a reference substance corresponding to a normal object and the two or more types of luminescent probes. A method for determining abnormalities in an object.

6. the step of generating a signal is a step of irradiating the two or more types of luminescent probes with excitation light, The step of acquiring the signal is a step of acquiring the luminance and / or chromaticity of light. The method for determining abnormality in an object according to claim 5.

7. Before the step of determining whether or not there is an abnormality, The method further comprises the steps of: obtaining a plurality of reference signals by allowing a plurality of reference substances to interact with the two or more types of luminescent probes; and subjecting the plurality of reference signals to machine learning to create a trained model. The method for determining abnormality in an object according to claim 5.

Images