Method, model and program for assisting disease diagnosis or prognosis prediction using lung or extracellular particle in breathing

Abstract

To provide a new method, model, program or the like for acquiring information usable for a disease diagnosis or prognosis prediction by using a sample taken from lungs or exhaled breath.SOLUTION: A method for assisting disease diagnosis or prognosis prediction including: acquiring (A) extracellular particle information on a population of extracellular particles collected from lungs or exhaled breath of a subject; the extracellular particle information including scattered light information on scattered light of irradiated light about each extracellular particle of the population of extracellular particles, and light emission information on a plurality of light emission whose peak wavelengths are mutually different, caused from a component of the extracellular particles or a labelling substance connected to the component due to the irradiated light; and (B) disease information on a state of a disease in the subject from each subject included in a group of subjects; and acquiring information for assisting the disease diagnosis or the prognosis prediction related to the extracellular particles collected from the lungs or the exhaled breath by machine learning on the basis of a data set including the extracellular particle information and the disease information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method, a model, a program, etc. for assisting in disease diagnosis or prognosis prediction using extracellular particles in the lungs or exhaled breath. [Background technology]

[0002] Extracellular particles ranging in size from several tens of nanometers to several micrometers containing biologically active molecules such as nucleic acids exist in living organisms, and their involvement in various diseases has attracted attention. For example, during influenza virus infection, in addition to viral particles, nucleic acids such as host cell-derived vesicles and neutrophil extracellular traps are released extracellularly and contribute to the pathogenesis of the disease. Therefore, quantification of these extracellular particles could be a new indicator for assessing the pathology of disease and the efficacy of vaccines and therapeutic drugs. However, while analytical methods at the protein and single-cell levels have been established, useful methods for analyzing extracellular particles at the single-particle level are limited.

[0003] In recent years, flow cytometry has attracted attention for single-particle analysis of nanoparticles, including extracellular vesicles. Because flow cytometry is a tool developed for cell analysis, it has been technically difficult to perform flow cytometric analysis of nanoparticles, which are 1 / 100 the size of a cell. As the detection sensitivity of flow cytometers has improved, size measurement based on scattered light and one- or two-color analysis using fluorescently labeled antibodies or low-molecular-weight fluorescent reagents have been reported (Non-Patent Documents 1 to 4).

[0004] Lung cancer is a leading cause of cancer-related deaths, and low-dose computed tomography (LDCT) is recommended as a routine screening tool for high-risk groups. [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Brittain, GCIV. et al. A Novel Semiconductor-Based Flow Cytometer with Enhanced Light-Scatter Sensitivity for the Analysis of Biological Nanoparticles. Sci. Rep. 9, 16039 (2019). [Non-patent document 2] Danielson, KM et al. Diurnal Variations of Circulating Extracellular Vesicles Measured by Nano Flow Cytometry. PLoS One 11, e0144678 (2016). [Non-patent document 3] van der Vlist, EJ et al. Fluorescent labeling of nano-sized vesicles released by cells and subsequent quantitative and qualitative analysis by high-resolution flow cytometry. Nat. Protoc. 7, 1311-1326 (2012). [Non-patent document 4] Morales-Kastresana, A. et al. Labeling Extracellular Vesicles for Nanoscale Flow Cytometry. Sci. Rep. 7, 1878 (2017). Summary of the Invention [Problem to be solved by the invention]

[0006] The analytical techniques reported so far can only obtain part of the information about the microparticles, for example, information about specific marker proteins or constituent molecules.

[0007] Furthermore, LDCT's invasiveness and high false-positive rate limit its use in screening low-risk groups. Therefore, a non-invasive method for detecting lung cancer that complements LDCT screening is needed. In particular, the histological type of lung cancer is important information for determining treatment strategies, but it is currently difficult to obtain without biopsy.

[0008] Therefore, an object of the present invention is to provide a novel method, model, program, etc. for obtaining information usable for diagnosing or predicting the prognosis of a disease using a sample collected from the lungs or exhaled breath. Another object of the present invention is to provide a novel method, model, program, etc. for obtaining information usable for non-invasively diagnosing or predicting the prognosis of a lung disease. [Means for solving the problem]

[0009] According to one embodiment of the present invention, a method for assisting in the diagnosis or prognosis of a disease includes obtaining, from each subject included in a population of subjects, (A) extracellular particle information regarding a population of extracellular particles collected from the lungs or exhaled breath of the subject, the extracellular particle information including, for each extracellular particle of the population of extracellular particles, scattered light information regarding scattered light of irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths resulting from the irradiated light and produced from components of the extracellular particles or labeled substances bound to the components; and (B) disease information regarding the state of the disease in the subject; and obtaining, by machine learning, information for assisting in the diagnosis or prognosis of a disease associated with the extracellular particles collected from the lungs or exhaled breath based on a dataset including the extracellular particle information and the disease information.

[0010] This method uses machine learning using extracellular particle information on a population of extracellular particles collected from the lungs or exhaled breath of a subject and disease information on the state of the disease in the subject, and by associating multiple characteristics, such as the size and components of the extracellular particles collected from the lungs or exhaled breath, with information on the state of the disease, it is possible to obtain information to assist in the diagnosis or prognosis prediction of diseases related to extracellular particles collected from the lungs or exhaled breath.

[0011] According to one aspect of the present invention, a method for analyzing a sample collected from the lungs or exhaled breath and containing a population of extracellular particles labeled with a plurality of labeling substances includes acquiring extracellular particle information for each extracellular particle contained in the population of extracellular particles, the information including scattered light information regarding scattered light of irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are caused by the irradiated light and that are generated from components of the extracellular particles or labeling substances bound to the components, and analyzing the population of extracellular particles by machine learning based on the extracellular particle information.

[0012] This method analyzes a population of extracellular particles collected from the lungs or exhaled breath of a subject using machine learning based on extracellular particle information including scattered light information and luminescence information, and therefore can analyze the characteristics of the population of extracellular particles taking into account multiple characteristics such as the size and constituent components of the extracellular particles. [Effects of the Invention]

[0013] According to the present invention, it is possible to provide novel methods, models, programs, etc. for obtaining information usable for diagnosing or predicting the prognosis of a disease using samples collected from the lungs or exhaled breath; or novel methods, models, programs, etc. for obtaining information usable for non-invasively diagnosing or predicting the prognosis of a lung disease. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a flowchart showing an example of a method for assisting in the diagnosis or prognosis prediction of a disease according to the present embodiment. [Figure 2] 1 is a flowchart showing an example of a method for assisting in the diagnosis of a disease according to the present embodiment. [Figure 3] 1 is a flowchart showing an example of a method for assisting prognosis prediction according to the present embodiment. [Figure 4] 1 is a flowchart showing an example of a method for analyzing extracellular particles according to the present embodiment. [Figure 5] 1 is a flowchart showing an example of information processing in the method for analyzing extracellular particles of the present embodiment. [Figure 6] 1 shows an example of the configuration of an analyzer for extracellular particles according to this embodiment. [Figure 7] FIG. 2 is a block diagram showing an example of the functional configuration of an information processing unit in the analyzer for extracellular particles of the present embodiment. [Figure 8] FIG. 2 is a block diagram showing an example of the physical configuration of an information processing unit in the analyzer for extracellular particles of the present embodiment. [Figure 9] BD Influx optimized forward scatter detection and sorting of extracellular particles. (A) Flow cytometry analysis and threshold trigger type for extracellular particle (EP) sorting. (B) Overview of the optical system of the BD Influx (high-resolution cell sorter) and the BD Aria III (conventional cell sorter). (C) A mixture of 100, 200, 500, and 1000 nm fluorescent beads was analyzed using FSC threshold trigger on the BD Influx (top left panel) or the BD Aria III (top right panel). Beads detected on the BD Influx (1000, 500, 200, and 100 nm) or the BD Aria III (1000 nm) were sorted and reanalyzed on the BD Influx (bottom panel). Pie charts show the percentage of each bead and the purity of the sorted beads (center of the chart). [Figure 10]Multiparametric analysis of extracellular particles and cells in BALF during influenza virus infection. (A-C) BALF was collected from mice uninfected with influenza virus or from mice 4 days post-infection. (B) EP in BALF was measured using a BD Influx™ system after staining with the indicated probes. (C) Cells and extracellular particles in BALF were measured using a BD Influx™ system without droplet formation. To analyze the morphology of extracellular particles, BALF was centrifuged at 400 x g, and the extracellular particles in the supernatant were observed by transmission electron microscopy (TEM) after negative staining. (D) All extracellular particles detected in (C) were analyzed by t-SNE. Each stained particle was visualized by the indicated color. The intensity of each channel is shown as a heat map. (E) Overview of the gating strategy used to exclude noise signals and unstained particles from subsequent t-SNE analysis. Stained particles of each color were gated and extracted using an "or" gate. (F) The results of t-SNE analysis of stained particles are shown in the same manner as in (D). (G-H) BALF from mice infected with different doses of influenza virus was stained with the indicated probes. Stained cells and extracellular particles were gated and analyzed by t-SNE. (G) Representative data from t-SNE analysis of cells (upper panel) and stained particles (lower panel) from each group are shown. Each cell population or stained particle was visualized by the color indicated in the figure. (H) The number of each particle and cell in the 2-minute measurement is shown (n = 3 per group). Statistical analysis was performed using one-way ANOVA with Tukey's multiple comparison test (*P < 0.05; **P < 0.01; ***P < 0.001). [Figure 11]Multiparametric analysis of EPs in human BALF for prognostic prediction of lung disease patients is shown. (A) The breakdown of human BALF samples is shown. Data from all samples (n=132) was analyzed using dimensionality reduction. Data from 46 patients whose prognosis could be tracked was used to generate a prognostic prediction model. CTD-ILD refers to connective tissue disease-associated interstitial pneumonia, AE-ILD refers to acute exacerbation of interstitial pneumonia, IPF refers to idiopathic pulmonary fibrosis, IIPs refers to idiopathic interstitial pneumonia, and FHP refers to fibrosing hypersensitivity pneumonitis. (B) Human BALF was stained with the indicated probes. Stained particles in 132 samples were analyzed using t-SNE and visualized using the indicated colors. (C) Conceptual diagram illustrating feature engineering of flow cytometry data for machine learning. To transform clinical sample EP data, which contains multidimensional parameters for each particle, while preserving dimensionality, binned data was created by separating EPs by FSC and fluorescence parameters. (D) PCA maps of 132 patients visualized using the 1490-bin EP dataset are shown. Patients with lung disease were divided into the main group and outliers based on the PCA maps. DILD refers to drug-induced lung disease. The vital capacity and survival curves of the two groups are shown (for vital capacity, the main group was n = 107, with outliers n = 3; for survival curve, the main group was n = 127, with outliers n = 5). (E) LightGBM was used to develop prognostic models using the clinical and / or EP datasets. The models were validated by LOOCV (n = 46). Statistical analysis was performed using Student's t-test or log-rank test for (D) and two-tailed Pearson correlation test for (E). [Figure 12]Multiparametric analysis of EP in human EBCs for the diagnosis of non-squamous lung cancer is shown. (A) Overview of human EBC sample analysis. (B) Human EBCs were stained with the indicated probes. Stained particles from 24 samples were analyzed using t-SNE and visualized with the colors indicated. (C) PCA maps of 24 donors (upper panel) and 22 donors (lower panel) visualized using the 174-bin EP dataset after removing two outliers. (D) A diagnostic model was developed using linear SVC (n = 24) using the 174-bin EP dataset. The model was validated using LOOCV. (E-F) Features for the diagnosis of non-squamous lung cancer were selected from the 174 bins using SelectKBest. (F) The top three clusters in the t-SNE plot shown in (B) are shown, along with the counts for each cluster. (G) Human EBCs were stained with the indicated probes. (H) A diagnostic model was constructed using linear SVC with the number of EpCAM- or podoplanin-positive particles (n = 24). The model was validated by look-up-and-convergence vein ... DETAILED DESCRIPTION OF THE INVENTION

[0015] Hereinafter, an embodiment of the present invention (hereinafter referred to as "the present embodiment") will be described in detail with reference to the drawings. However, the present invention is not limited to this embodiment, and various modifications are possible without departing from the spirit of the present invention. In the following description of the drawings, the same or similar parts are denoted by the same or similar reference numerals. The drawings are schematic and do not necessarily correspond to actual dimensions, proportions, etc. Parts in which the dimensional relationships and proportions differ from one another may be included.

[0016] [Method for assisting in diagnosis or prognosis of disease] The method for assisting in the diagnosis or prognosis of a disease of this embodiment includes obtaining, from each subject included in a population of subjects, (A) extracellular particle information regarding a population of extracellular particles collected from the lungs or exhaled breath of the subject, the extracellular particle information including, for each extracellular particle of the population of extracellular particles, scattered light information regarding scattered light of irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are generated from components of the extracellular particles or labeled substances bound to the components due to the irradiated light; and (B) disease information regarding the state of the disease in the subject; and obtaining, by machine learning, information for assisting in the diagnosis or prognosis of a disease associated with the extracellular particles collected from the lungs or exhaled breath based on a dataset including the extracellular particle information and the disease information.

[0017] As demonstrated in the examples described below, this method uses machine learning to associate the characteristics of extracellular particles collected from the lungs or exhaled breath with information on the disease state through extracellular particle information on a population of extracellular particles collected from the lungs or exhaled breath of a subject and disease information on the disease state of the subject, thereby making it possible to obtain information to assist in the diagnosis or prognosis prediction of diseases related to extracellular particles collected from the lungs or exhaled breath.

[0018] More specifically, this method obtains multiple characteristics, such as the size and components, of each extracellular particle collected from the lungs or exhaled breath of a subject at the single particle level, and by associating the obtained data with disease information and performing machine learning, it is possible to obtain information to assist in the diagnosis or prognosis prediction of a disease, such as characteristics of extracellular particles that are characteristic of the patient, new disease markers, disease diagnostic models, and prognosis prediction models.

[0019] It has not been possible to diagnose or predict the prognosis of, for example, a disease related to the trachea, bronchi, or lungs based on a population of extracellular particles collected from the lungs or exhaled breath of a subject. The present disclosure is the first to demonstrate that disease diagnosis and prognosis prediction are possible by analyzing a population of extracellular particles collected from the lungs or exhaled breath of a subject, and it is particularly surprising that lung disease can be diagnosed and predicted noninvasively using a population of extracellular particles collected from exhaled breath.

[0020] Although methods for multicolor staining cells and analyzing them based on multiple pieces of information are known, the present embodiment is a completely different analytical technique from such multicolor cell analysis. Specifically, extracellular microparticles have a surface area approximately 1 / 100 to 1,000,000 times smaller than cells, and a volume approximately 1 / 1,000 to 1,000,000,000 times smaller than cells. Therefore, due to the bulkiness of the staining reagents, multicolor staining of extracellular microparticles using methods similar to those used for cells is difficult. Therefore, multicolor analysis of extracellular microparticles has not been performed in existing prior art, including the above-mentioned non-patent literature, and specific marker proteins have only been analyzed by staining one or two components. In other words, there have been very few attempts to comprehensively analyze extracellular microparticles at the single-particle level based on multiple pieces of information. The present embodiment enables comprehensive analysis of extracellular microparticles at the single-particle level by visualizing multiple characteristics of extracellular microparticles.

[0021] FIG. 1 is a flowchart showing an example of a method for assisting in the diagnosis or prognosis of a disease according to this embodiment. Hereinafter, one aspect of the method for assisting in the diagnosis or prognosis of a disease according to this embodiment will be described with reference to the drawings as appropriate. Note that FIG. 1 and the following description are one aspect of this embodiment, and are not intended to limit the present invention. For example, the method of this embodiment may not include a step of preparing a sample. In that case, for example, a sample containing a population of extracellular microparticles in which the components of the extracellular microparticles have already been labeled may be obtained and analyzed.

[0022] (Sample preparation) 1, first, a sample to be used in the method of this embodiment is prepared (S110). The sample contains a population of extracellular particles collected from the lungs or exhaled breath of a subject. Examples of such samples include bronchoalveolar lavage fluid (BALF) and exhaled breath coagulate (EBC). However, EBC is preferred from the viewpoint of greater non-invasiveness.

[0023] BALF can be collected by conventional methods, for example, by using a bronchoscope to inject physiological saline into the bronchus or bronchioles of the middle lobe or lingule, followed by suction. After collection, BALF may be used after removing mucus and cellular components by filtration or centrifugation. EBCs can be collected using an EBC collection device such as an R tube (Respiratory Research Inc.) by breathing quietly for 3 to 20 minutes or 5 to 15 minutes.

[0024] In the analysis sample, the extracellular particles contain a component that emits light upon irradiation with light, or if they do not contain such a component, they contain a component to which a labeling substance that emits light upon irradiation with light is bound. Therefore, it is preferable that the sample preparation includes binding a labeling substance to the component of the extracellular particles. Examples of the component that emits light upon irradiation with light include components that generate autofluorescence, such as melanin.

[0025] As used herein, "extracellular microparticles" refer to particles with a size of several tens of nanometers to several micrometers that are released outside cells, including extracellular vesicles such as exosomes, microvesicles, and apoptotic bodies, as well as extracellular particles such as protein-nucleic acid complexes. The size of the extracellular microparticles may be, for example, 1.0 nm to 50 μm, 10 nm to 10 μm, or 50 nm to 5.0 μm.

[0026] As used herein, the term "labeling substance" refers to a substance that specifically binds to a specific component of extracellular microparticles and can be optically detected by light irradiation. Here, the binding mode of the labeling substance to the component is not particularly limited, and examples include covalent bonds, bonds due to hydrophobic interactions, hydrogen bonds, and ionic bonds. The labeling substance may be a substance that generates scattered light when irradiated with light, or a substance that generates luminescence, and in particular, a substance that generates fluorescence. The labeling substance may be, for example, a staining reagent for a component of the extracellular microparticles.

[0027] As used herein, "staining" means that a molecule that emits light (typically emits fluorescence) upon light irradiation is specifically chemically or physically bound to a specific component of an extracellular microparticle, and that a molecule that emits light (typically emits fluorescence) upon light irradiation is specifically chemically or physically bound to a specific component of an extracellular microparticle. Furthermore, a "staining reagent" is a substance for specifically staining a specific component of an extracellular microparticle, i.e., a substance that specifically binds to a specific component of an extracellular microparticle and emits light (typically emits fluorescence) upon light irradiation.

[0028] The components of the extracellular particles to be labeled with a labeling substance preferably include at least one of nucleic acids, proteins, lipids, and sugar chains, more preferably at least three of nucleic acids, proteins, lipids, and sugar chains, and even more preferably all of nucleic acids, proteins, lipids, and sugar chains. By labeling and detecting these components, a population of extracellular particles can be analyzed with higher precision. In the analysis sample, constituent molecules other than nucleic acids, proteins, lipids, and sugar chains may be labeled with a labeling substance.

[0029] In the analysis sample, the extracellular microparticles preferably have three or more constituent components labeled with labeling substances, and more preferably four or more constituent components labeled with labeling substances. Furthermore, the extracellular microparticles may have constituent components labeled with preferably three or more, more preferably four or more, even more preferably five or more, even more preferably six or more, even more preferably seven or more, and particularly preferably eight or more labeling substances. The upper limit of the number of types of labeling substances contained in the population of extracellular microparticles is not particularly limited, and may be, for example, 40, 30, 20, 16, 15, 12, or 10.

[0030] In this case, in the population of extracellular microparticles, each of the nucleic acids, proteins, lipids, and sugar chains may be stained with one or more staining reagents. Each type of nucleic acid, protein, lipid, and sugar chain may be stained with 1 to 10, 1 to 8, 1 to 5, 2 to 10, 2 to 8, or 2 to 5 staining reagents.

[0031] Furthermore, the population of extracellular microparticles is preferably stained with three or more staining reagents, more preferably four or more, even more preferably five or more, even more preferably six or more, even more preferably seven or more, and particularly preferably eight or more staining reagents. The upper limit of the number of types of staining reagents contained in the population of extracellular microparticles is not particularly limited, but may be, for example, 40, 30, 20, 16, 15, 12, or 10.

[0032] As described above, extracellular microparticles have an extremely small volume and surface area compared to cells. Therefore, from the viewpoint of labeling extracellular microparticles with multiple labeling substances, the labeling substance or staining reagent in the extracellular microparticles preferably contains a substance other than an antibody, more preferably contains a compound with a molecular weight of 20,000 or less, and even more preferably contains a low molecular weight compound. For example, the extracellular microparticles may contain a compound or low molecular weight compound with a molecular weight of 20,000 or less as a labeling substance or staining reagent for nucleic acids and / or lipids. The molecular weight of the low molecular weight compound may be, for example, 500 or less, or 400 or less, or may be 100 or more, or 200 or more. The labeling substance or staining reagent in the extracellular microparticles may contain a medium molecular weight compound with a molecular weight of 500 or more and 20,000 or less.

[0033] The nucleic acid staining reagent is not particularly limited as long as it is a substance that can stain nucleic acids, and examples include those shown in the table below. The nucleic acid to be stained may be selected depending on the sample and the purpose of analysis, etc. Preferred nucleic acid staining reagents are those whose binding mode is "Bis-intercalator" and those whose binding mode is "Minor groove binder." Of these, those whose binding mode is "Bis-intercalator" and those whose binding mode is "Minor groove binder" in the table below are more preferred, and it is even more preferred to use a combination of a staining reagent whose binding mode is "Bis-intercalator" and a staining reagent whose binding mode is "Minor groove binder" in the table below.

[0034] [Table 1]

[0035] [Table 2]

[0036] Lipid staining reagents are not particularly limited as long as they can stain the interior of lipids or lipid particles. Examples include amphipathic intercalator-type staining reagents such as PKH67, DiA, DiB, Neuto-DiO, DiD, and CellMask (CellMask Green, CellMask Orange, and CellMask Deep Red); amine-reactive staining reagents that fluoresce upon decomposition by esterases in lipid bilayers, such as 5-carboxyfluorescein diacetate succinimidyl ester (CFSE), 6-carboxyfluorescein diacetate succinimidyl ester (CFSE), and cell trace violet (CTV); and fluorescently labeled proteins that bind to phosphatidylserine (PS), such as annexin V. The lipid to be stained may be selected depending on the sample and the purpose of analysis. The lipid staining reagent preferably contains an amphipathic intercalator-type staining reagent, more preferably contains CellMask, and even more preferably contains CellMask Deep Red.

[0037] The protein staining reagent is not particularly limited as long as it can stain proteins, and examples thereof include fluorescently labeled antibodies and staining reagents for assessing cell viability, such as Live / Dead stain Near-IR and Live / Dead fixable dead cell stain. The protein to be stained may be selected depending on the sample and the purpose of analysis.

[0038] The staining reagent for glycans is not particularly limited as long as it is a substance that can stain glycans, and examples thereof include fluorescently labeled lectins such as CF405M WGA and CF568 PNA. The glycans to be stained may be selected depending on the sample, the purpose of the analysis, etc.

[0039] When staining extracellular particles, staining reagents with different peak wavelengths are selected from the above staining reagents, which allows for a one-to-one correspondence between the luminescence detected in the optical detection described below and the staining reagent, thereby enabling more accurate analysis of the extracellular particle population.

[0040] The sample preparation step S110 preferably includes, for example, labeling a sample containing a population of extracellular particles with multiple labeling substances that bind to components of the extracellular particles, and staining with at least three types selected from the group consisting of a nucleic acid staining reagent, a protein staining reagent, a lipid staining reagent, and a sugar chain staining reagent. The number of types of labeling substances or staining reagents used, as well as examples of labeling substances and staining reagents, are as described above. For example, during sample preparation, it is preferable to stain a sample containing a population of extracellular particles with all of a nucleic acid staining reagent, a protein staining reagent, a lipid staining reagent, and a sugar chain staining reagent.

[0041] The method for labeling a population of extracellular microparticles with a labeling substance or staining reagent may be any known method appropriate for the labeling substance or staining reagent used and the object to be labeled or stained. Furthermore, components of extracellular microparticles, such as nucleic acids, proteins, lipids, and sugar chains, may be stained simultaneously or stepwise. Furthermore, when two or more labeling substances or staining reagents are used for each of the components of extracellular microparticles, such as nucleic acids, proteins, lipids, and sugar chains, the microparticles may be labeled or stained simultaneously with these two or more labeling substances or staining reagents, or stepwise. When labeling or staining is performed stepwise, the population of extracellular microparticles may be washed between labeling or staining with one labeling substance or staining reagent and labeling or staining with a second labeling substance or staining reagent; however, it is preferable not to perform such washing in order to prevent the outflow of extracellular microparticles, particularly extracellular microparticles containing nucleic acids.

[0042] An example of a sample preparation method will be described below, taking the case of staining nucleic acids, proteins, lipids, and sugar chains for a population of extracellular microparticles as an example. However, it goes without saying that the sample preparation method in this embodiment is not limited to the following example. First, a sample containing a population of extracellular microparticles is optionally diluted or concentrated to an appropriate concentration. Then, a lipid staining reagent, such as CellMask, is added and allowed to stand, for example, at a temperature of 0°C to 30°C for 1 minute to 30 minutes. Next, the remaining nucleic acid staining reagent, protein staining reagent, and sugar chain staining reagent are added and allowed to stand, for example, at a temperature of 0°C to 30°C for 5 minutes to 60 minutes. In this way, by performing staining with a lipid staining reagent in advance, the lipid staining efficiency tends to be improved.

[0043] (Acquisition of extracellular particle information) 1, next, extracellular particle information on a population of extracellular particles is obtained using a sample prepared for each subject (S111). The extracellular particle information is information including, for each extracellular particle, scattered light information on scattered light of irradiated light and luminescence information on a plurality of luminescences with different peak wavelengths that are caused by the irradiated light and are generated from components of the extracellular particles or labeling substances bound to the components. In step S111, extracellular particle information including scattered light information and luminescence information of each extracellular particle of the population of extracellular particles corresponding to each subject is linked and obtained.

[0044] In the extracellular particle information acquisition step S111, a group of extracellular particles is irradiated with light, and scattered light information and luminescence information are acquired from each extracellular particle.

[0045] The method of irradiating light is not particularly limited as long as scattered light and luminescence can be detected for each particle from a group of extracellular particles during light detection, as described below. For example, light may be irradiated onto a group of extracellular particles all at once, or light may be irradiated onto each particle or several particles (e.g., about 1 to 10 particles) at a time while the extracellular particles are flowing through a flow channel. Here, one or more types of light may be irradiated onto each particle.

[0046] Furthermore, the irradiated light may be white light or monochromatic light. When irradiating monochromatic light, the wavelength of the irradiated light may be selected according to the wavelength of the excitation light of the constituent components of the extracellular microparticles or the labeling substance bound to the constituent components. For example, when a population of extracellular microparticles is labeled with multiple types of labeling substances that emit excitation light with different wavelengths, multiple types of monochromatic light corresponding to the wavelengths of the excitation light may be irradiated.

[0047] The method of irradiating a group of extracellular microparticles with light particle by particle is not particularly limited, but may be, for example, a method using a flow cytometer. The irradiated light may be a monochromatic laser. The wavelength of the irradiated light is not particularly limited, but may be, for example, in the range of 400 to 600 nm.

[0048] By irradiating the extracellular microparticles with light as described above, each extracellular microparticle generates scattered light and luminescence (typically fluorescent light) derived from the constituents of the extracellular microparticles or the labeling substances bound to the constituents. In this embodiment, the scattered light and luminescence are detected for each extracellular microparticle, thereby obtaining scattered light information and luminescence information. In this embodiment, the sample is prepared or prepared so that, in this light detection, multiple luminescences with different peak wavelengths are detected from the constituents of the extracellular microparticles or the labeling substances bound to the constituents.

[0049] Examples of scattered light to be detected include forward scattered light and side scattered light. Furthermore, in addition to or instead of normal scattered light, a polarized component of scattered light may be detected. To broaden the range of detectable extracellular particle sizes, for example, forward scattered light (FSC) and polarized (e.g., vertically polarized) forward scattered light (FSC-perp) may be detected simultaneously. At least two or more of side scattered light (SSC), forward scattered light (FSC), and polarized forward scattered light (FSC-perp) may be detected simultaneously, or all of them may be detected simultaneously.

[0050] The number of luminescences derived from the extracellular particles detected is equal to the number of types of constituent components that generate autoluminescence plus the number of types of labeling substances contained in the extracellular particles. That is, in this embodiment, multiple luminescences with different peak wavelengths are detected.

[0051] An appropriate photodetector may be used to detect scattered light and luminescence. One or more photodetectors may be used. When one photodetector is used, a spectrum analyzer may be used to detect multiple luminescence beams with different peak wavelengths. When more than one photodetector is used, an appropriate spectroscopic element may be used to separate the scattered light and luminescence beams generated by the above-mentioned light irradiation into their respective peak wavelengths, and each light beam may be detected by a detector corresponding to each light beam.

[0052] Examples of the photodetector include a photomultiplier tube and a photodiode, and the photodetector may be one that is built into a flow cytometer.

[0053] The scattered light information is information reflecting the size of extracellular particles, and may be information about scattered light detected as described above, but is preferably scattered light intensity. The scattered light information preferably includes any of forward scattered light (FSC) intensity, polarized (e.g., vertically polarized) forward scattered light (FSC-perp) intensity, and side scattered light (SSC) intensity, and may include at least two or more of these. The scattered light information may include both forward scattered light (FSC) intensity and polarized (e.g., vertically polarized) forward scattered light (FSC-perp) intensity.

[0054] The luminescence information may be information relating to the plurality of luminescences detected as described above, but is preferably the luminescence intensity relating to the plurality of luminescences. The luminescence information may include the light intensity at a plurality of specific wavelengths detected from each extracellular microparticle. For example, the luminescence information may be the detected light intensity at the fluorescent wavelength of each staining reagent used in the sample preparation step S110.

[0055] (Acquisition of disease information) 1, in parallel with steps S110 and S111, or before or after these steps, disease information relating to the disease state of each subject for which extracellular particle information has been acquired is acquired (S12). Note that the extracellular particle information acquired in the extracellular particle information acquisition step S111 and the disease information acquired in the disease information acquisition step S12 are acquired in association with each subject.

[0056] The disease information is not particularly limited as long as it is information about the state of the disease of the subject, and examples thereof include whether or not the subject is suffering from a specific disease, whether or not the subject is suffering from any disease selected from a specific disease group, what kind of disease the subject is suffering from, whether or not the subject is healthy, the severity of the specific disease, the prognosis of the disease, etc. The disease information may be obtained from clinical data of the subject.

[0057] In this embodiment, the diseases for which disease information is obtained are not particularly limited as long as they can affect extracellular microparticles collected from the lungs or exhaled breath, and may be, for example, respiratory diseases such as lung diseases, bronchial diseases, and tracheal diseases, and more specifically, include lung tumors, pulmonary tuberculosis, nontuberculous mycobacterial disease, interstitial lung diseases, pulmonary fibrosis, and pneumonia.

[0058] (Acquisition of information to assist in disease diagnosis or prognosis prediction) Next, in FIG. 1, the extracellular particle information obtained in step S111 and the disease information obtained in step S12 are used as a data set to obtain information to assist in the diagnosis or prognosis prediction of diseases associated with extracellular particles collected from the lungs or exhaled breath through machine learning (S13).

[0059] The information to be acquired is not particularly limited as long as it is information that assists in the diagnosis or prognosis prediction of a disease, and examples include characteristics of extracellular microparticles that are characteristic of patients with a particular disease, new markers for the disease, diagnostic models for the disease, and prognosis prediction models.

[0060] To obtain such information, machine learning, for example, supervised learning, is performed using a dataset containing extracellular particle information and disease information. In step S13, machine learning is performed using disease information as a response variable and extracellular particle information as an explanatory variable, thereby making it possible to obtain information for assisting in the diagnosis or prognosis prediction of disease from the extracellular particle information. An example of such machine learning is described below.

[0061] (obtaining information to assist in disease diagnosis) Fig. 2 is a flowchart showing an example of a method for assisting disease diagnosis according to this embodiment. In the method of the embodiment shown in Fig. 2, sample preparation is performed for a first group of subjects having first disease information and a second group of subjects having second disease information, and extracellular particle information and disease information are acquired (S110A, B, S111A, B, S12A, B). The sample preparation steps S110A and B, the extracellular particle information acquisition steps S111A and B, and the disease information acquisition steps S12A and B may be performed in the same manner as steps S110, S111, and S12, respectively, described using Fig. 1.

[0062] Next, machine learning is performed based on the acquired information to acquire information to assist in disease diagnosis (S131). Specifically, for the first and second populations, feature selection is performed using disease information as a response variable and extracellular particle information as an explanatory variable, thereby making it possible to extract characteristic features of extracellular particles in the first or second population.

[0063] For example, when the first population is a population of subjects suffering from a specific disease and the second population is a population of subjects not suffering from the disease, this method can be used to obtain information on extracellular particles characteristic of a population of extracellular particles collected from the lungs or exhaled breath of patients suffering from the specific disease. Obtaining such information can aid in the diagnosis of the disease, for example, by collecting a population of extracellular particles from the lungs or exhaled breath of a subject and confirming the concentration of the characteristic extracellular particles. Furthermore, if the characteristic extracellular particles are found to have a specific component, the component can be identified as a marker for the disease.

[0064] The algorithm for feature selection is not particularly limited, and for example, algorithms based on the filter method, wrapper method, embedded method, etc. can be used, and may be selected appropriately depending on the size of the data set and the sizes of the extracellular particle information and disease information. Feature selection may be performed, for example, by statistically analyzing and scoring the influence of each explanatory variable on the objective variable, and selecting the top explanatory variables as features.

[0065] Alternatively, this method can also be used to create a diagnostic model for a specific disease. For example, a diagnostic model that estimates disease information from extracellular particle information can be created by generating a model through machine learning using training data, which is a dataset for a first population and a second population, in which disease information is the objective variable and extracellular particle information is the explanatory variable. To generate the model, techniques such as convolutional neural networks (CNNs), decision trees, random forests, gradient boosting, naive Bayes methods, and support vector machines may be used, or techniques such as ensemble learning that combine these techniques may be used. Furthermore, information on extracellular particles characteristic of a population of extracellular particles collected from the lungs or exhaled breath of a patient suffering from a specific disease may be obtained using the above method, and a model may be created using these characteristics as explanatory variables.

[0066] In selecting the first and second populations, extracellular particles of all subjects may be analyzed in advance using the analysis method of this embodiment described below, and a population that exhibits characteristics in the extracellular particle information may be selected, or two or more populations classified into different clusters may be selected.

[0067] 2 shows an example of classification into a total of two groups, a first group and a second group, but it goes without saying that the above processing may be performed for groups of three or more subjects. When classifying into three or more groups of subjects, classification may be performed for a plurality of diseases, for example, into a first group consisting of subjects suffering from a first disease, a second group consisting of subjects suffering from a second disease, and a third group consisting of subjects suffering from neither the first disease nor the second disease.

[0068] (Acquisition of information to assist in prognosis prediction) Fig. 3 is a flowchart showing an example of a method for assisting in prognosis prediction according to this embodiment. In the method of the embodiment shown in Fig. 3, sample preparation is performed for a patient with a disease, and extracellular microparticle information and disease information are obtained (S110, S111, S12). These steps may be performed in the same manner as steps S110, S111, and S12, respectively, described with reference to Fig. 1. However, in step S12, the prognosis status of each patient is obtained.

[0069] Next, machine learning is performed based on the acquired information to acquire information to assist in prognosis prediction (S132). Specifically, by using disease information (prognosis state) as the objective variable and extracellular particle information as the explanatory variable, it is possible to acquire the relationship between the extracellular particle information and the prognosis state after a certain period of time has passed since the extracellular particle information was acquired, or to predict prognostic factors. By acquiring such information, it is possible to predict the prognosis state after a certain period of time, for example, by collecting and analyzing a group of extracellular particles from the patient's lungs or exhaled breath.

[0070] This method can also be used to create a prognosis prediction model for a patient. For example, a prognosis prediction model can be created that estimates the prognosis after a certain period of time from the extracellular particle information by generating a model through machine learning using a dataset in which disease information (prognosis state) is used as the objective variable and extracellular particle information is used as the explanatory variable for the first and second populations. To generate the model, techniques such as convolutional neural networks (CNNs), decision trees, random forests, logistic regression, gradient boosting, and support vector machines may be used, or techniques such as ensemble learning that combine these may be used. Furthermore, a model may be generated using the prognostic factors obtained by the above-mentioned method as explanatory variables.

[0071] (Data preprocessing) As described above, in the method of this embodiment, machine learning is performed using extracellular particle information, which is high-dimensional vector information including scattered light information and luminescence information. Therefore, in this embodiment, it is preferable to process the extracellular particle information before performing machine learning. Below, an example of processing the extracellular particle information performed by an information processing device will be described.

[0072] In the above-mentioned step S111, for example, by light irradiation and light detection, the scattered light intensity and the emission intensity of each of a plurality of emission lights having different peak wavelengths are obtained for each extracellular particle in the group of extracellular particles. For example, when the group of extracellular particles is stained with three kinds of staining reagents having emission wavelengths of λ1, λ2, and λ3, for each extracellular particle, for example, the intensity I of the forward scattered light is obtained. FSC , the intensity of polarized forward scattered light I p-FSC , the intensity of side scattered light I SSC , and information on the intensities Iλ1, Iλ2, and Iλ3 of three types of luminescence with wavelengths λ1, λ2, and λ3, respectively, is obtained. That is, for each extracellular particle, a vector is obtained whose components are the intensity of scattered light and the intensities of multiple luminescences. Explaining the above example, for each extracellular particle, the vector (I FSC ,I p-FSC ,I SSC,Iλ1,Iλ2,Iλ3) are obtained.

[0073] That is, the information processing device generates a vector whose components are the detected scattered light intensity and the emission intensities of the plurality of emitted light beams having different peak wavelengths. The information processing device then performs appropriate processing on the vector to reduce the data. Examples of such processing include a process that deletes vector components, a binning process that converts continuous values ​​into discrete values, or a normalization process, which do not substantially change the magnitude relationship between the components in the vector before processing. A process that does not substantially change the magnitude relationship between the components in the vector before processing refers to a process that does not change the magnitude relationship between the components in the vector before processing, or a process that discretizes the components in the vector before processing.

[0074] Generally, the scattered light intensity is information reflecting the size of the extracellular particles, and the luminescence intensity is information indicating whether or not they are labeled with a labeling substance. Therefore, when reducing the data, the data may be classified into several sizes based on the scattered light intensity, and the presence or absence of staining may be classified based on the luminescence intensity of each wavelength. In the above example, for example, the vector (I FSC ,I p-FSC ,I SSC ,Iλ1,Iλ2,Iλ3) to (I SC_bin ,Iλ 1_bin ,Iλ 2_bin ,Iλ 3_bin ) where I SC_bin is a discrete value that classifies the size of extracellular particles based on the scattered light intensity, and Iλ 1_bin , Iλ 2_bin , and Iλ 3_bin may be discrete values ​​that represent whether or not the sample is stained by each of the types of staining reagents.

[0075] Here, the information processing device may use vectors corresponding to all extracellular particles in the population of extracellular particles for machine learning, or may use vectors corresponding to a portion of a plurality of extracellular particles included in the population of extracellular particles for machine learning. This selection may be performed, for example, based on the components of the vectors of each extracellular particle.

[0076] For example, extracellular particles for which all of the components corresponding to the multiple emissions of the corresponding vectors are below the threshold may be excluded from the training data as noise components that are not stained at all. In this way, by excluding extracellular particles that are not labeled or stained, or that are only weakly labeled or stained, machine learning can be performed with higher accuracy.

[0077] The threshold value in the binning process may be set by the user of the information processing device, or may be set by the information processing device based on a predetermined algorithm. For example, a population of unlabeled or unstained extracellular microparticles and a population of labeled or stained extracellular microparticles may be prepared from the same population of extracellular microparticles, and the threshold value may be set based on the luminescence intensity contained in the luminescence information of the unlabeled or unstained extracellular microparticle population. In this case, the threshold value may be, for example, the maximum value, average value, mode value, average value + 2σ value, average value + 3σ value, or average value + 4σ value of the luminescence intensity in the population of unlabeled or unstained extracellular microparticles.

[0078] [Disease diagnosis model / prognosis prediction model] One aspect of this embodiment provides a trained model for diagnosing a disease or predicting a patient's prognosis, created by the above-described method. The disease is not particularly limited as long as it can affect extracellular particles collected from the lungs or exhaled breath, and may be, for example, a respiratory disease such as a lung disease, a bronchial disease, or a tracheal disease. More specifically, examples of the disease include lung tumors, pulmonary tuberculosis, nontuberculous mycobacterial disease, interstitial lung disease, pulmonary fibrosis, and pneumonia.

[0079] Such a trained model may be, for example, a model that diagnoses or predicts the prognosis of a subject's disease based on extracellular particle information obtained from extracellular particles collected from the subject's lungs or exhaled breath.

[0080] The trained model may be, for example, a model capable of diagnosing or predicting the prognosis of non-squamous lung cancer using extracellular particles in exhaled breath condensate. The non-squamous lung cancer diagnosis or prognosis prediction model may receive the above-described extracellular particle information regarding a population of extracellular particles collected from a subject's lungs or exhaled breath, and perform a diagnosis or prognosis prediction based on the concentration of at least one of EpCAM (epithelial cell adhesion molecule)-positive extracellular particles and podoplanin-positive extracellular particles in the population of extracellular particles. Furthermore, such a trained model may perform a diagnosis or prognosis prediction based on the concentration of extracellular particles positive for a cell membrane marker such as CellMask and / or a phosphatidylserine marker such as Annexin V.

[0081] Such a learning model may, for example, in the above-mentioned sample preparation step S110, stain a population of extracellular microparticles collected from the subject's lungs or exhaled breath for at least one selected from the group consisting of EpCAM, podoplanin, cell membrane, and phosphatidylserine, preferably EpCAM and podoplanin or cell membrane and phosphatidylserine, more preferably EpCAM and podoplanin.

[0082] Such a learning model may be created by the above-mentioned method, for example, by supervised learning using extracellular particle information obtained from extracellular particle samples collected from the lungs or exhaled breath of a subject with a predetermined disease and extracellular particle information obtained from extracellular particle samples collected from the lungs or exhaled breath of a subject without the disease. The method for generating the model is as described above.

[0083] [Method for analyzing extracellular particles] The method for analyzing extracellular particles of this embodiment is a method for analyzing a sample collected from the lungs or exhaled breath and containing a population of extracellular particles labeled with a plurality of labeling substances, and includes the steps of: acquiring extracellular particle information for each extracellular particle contained in the population of extracellular particles, the information including scattered light information regarding scattered light of irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are generated from the constituent components of the extracellular particles or labeling substances bound to the constituent components due to the irradiated light; and analyzing the population of extracellular particles by machine learning based on the extracellular particle information.

[0084] The analytical method of this embodiment acquires multiple characteristics of each extracellular particle, such as its size and constituent components, at the single particle level, and analyzes the obtained data using information processing, thereby analyzing the extracellular particles based on multiple pieces of information. This makes it possible to identify extracellular particles that serve as biomarkers for disease onset, disease cause, and / or prognosis prediction from among the various extracellular particles contained in biological samples. It is also possible to evaluate the characteristics of environmental particles or various nanoparticles contained in pharmaceuticals, etc. at the single particle level.

[0085] Figure 4 is a flowchart showing an example of the analytical method of this embodiment. The analytical method of this embodiment will be described below with reference to the drawings as appropriate. Note that the analytical method in Figure 4 and the following description is an example of the analytical method of this embodiment, and is not intended to limit the present invention. For example, the analytical method of this embodiment may not include a step of preparing a sample. In that case, for example, a sample containing a population of extracellular microparticles whose constituent components have already been labeled may be obtained and analyzed.

[0086] In Fig. 4, the sample preparation step S20 may be performed in the same manner as the sample preparation step S110 shown in Fig. 1. Furthermore, the extracellular particle information acquisition step S21 may be performed in the same manner as the extracellular particle information acquisition step S111 shown in Fig. 1.

[0087] (Analysis of extracellular particle populations) 4, the extracellular particle information acquired in step S111 is used to analyze the population of extracellular particles collected from the lungs or exhaled breath by machine learning (S23). Such analyses include analyses that identify characteristics of the population of extracellular particles, and specifically include visualization of the population of extracellular particles and clustering of the population of extracellular particles.

[0088] For example, in one embodiment, analyzing a population of extracellular particles may include: generating a feature vector for each extracellular particle contained in the population of extracellular particles based on the acquired extracellular particle information; mapping at least some of the extracellular particles contained in the population of extracellular particles in two or three dimensions based on the similarity of a representative vector consisting of some or all of the components of the feature vector, and labeling each of the mapped extracellular particles based on multiple components of the representative vector that correspond to at least the luminescence information, thereby creating a first scatter plot. According to this aspect, a population of extracellular particles collected from the lungs or exhaled breath is mapped in two or three dimensions based on the similarity of the vector generated based on the scattered light information and the luminescence information, so that the distribution of the similarity of the population of extracellular particles can be visualized taking into account multiple features such as size and constituent components.

[0089] This embodiment may further include classifying the mapped extracellular particles into two or more clusters based on the first scatter plot. According to this embodiment, it is possible to appropriately cluster a population of extracellular particles collected from the lungs or exhaled breath.

[0090] In addition, in this aspect, it may further comprise mapping at least a part of the plurality of extracellular particles contained in the population of extracellular particles in the same manner as the first scatter diagram, and visualizing one of the plurality of components corresponding to the detected plurality of luminescence of the component of the representative vector in each of the mapped plurality of extracellular particles, thereby creating a second scatter diagram.By creating the second scatter diagram as described above, it is possible to identify, for the plurality of extracellular particles mapped on the first scatter diagram, for example, extracellular particles that contain a large amount or hardly contain a predetermined component, and the cluster to which the extracellular particles belong.

[0091] Furthermore, this aspect may further include displaying the first and second scatter plots; and identifying characteristics of at least one of the clusters based on the displayed first and second scatter plots. According to this aspect, the user can compare the first scatter plot with the second scatter plot, confirm the content of a predetermined component in the extracellular particles contained in at least one of the clusters in the first scatter plot, and identify in more detail the type of extracellular particles contained in the cluster. Note that displaying the first and second scatter plots may be performed by the extracellular particle analysis device and / or an information processing unit provided therein, and identification of the characteristics may be performed by the extracellular particle analysis device and / or an information processing unit provided therein, and / or a user.

[0092] Furthermore, in one embodiment, analyzing the population of extracellular particles may include: generating a feature vector for each extracellular particle contained in the population of extracellular particles based on the acquired extracellular particle information; and classifying the extracellular particles contained in the population of extracellular particles into two or more clusters based on the similarity of representative vectors consisting of some or all of the components of the feature vector. According to this aspect, the population of extracellular particles collected from the lungs or exhaled breath is clustered based on the similarity of vectors generated based on scattered light information and luminescence information, so that the population of extracellular particles can be classified based on multiple features such as size and constituent components.

[0093] An example of information processing in the extracellular particle group analyzing step S23 will be described below with reference to FIG.

[0094] 5, a feature vector and a representative vector are created based on the extracellular particle information acquired in step S21 (S231). As explained in step S111, in step S21, for example, by light irradiation and light detection, the scattered light intensity and the emission intensity of each of a plurality of emissions having different peak wavelengths are obtained for each extracellular particle in the group of extracellular particles. For example, when the group of extracellular particles is stained with three kinds of staining reagents having emission wavelengths λ1, λ2, and λ3, for each extracellular particle, for example, the forward scattered light intensity I FSC , the intensity of polarized forward scattered light I p-FSC , the intensity of side scattered light I SSC , and information on the intensities Iλ1, Iλ2, and Iλ3 of three types of luminescence with wavelengths λ1, λ2, and λ3, respectively, is obtained. That is, for each extracellular particle, a vector is obtained whose components are the intensity of scattered light and the intensities of multiple luminescences. Explaining the above example, for each extracellular particle, the vector (I FSC ,I p-FSC ,I SSC ,Iλ1,Iλ2,Iλ3) are obtained.

[0095] In step S231, the information processing device generates a feature vector corresponding to each extracellular microparticle based on the detected scattered light intensity and the respective emission intensities of a plurality of emission lights having mutually different peak wavelengths. The feature vector may be a vector whose components are the detected scattered light intensity and the respective emission intensities of a plurality of emission lights having mutually different peak wavelengths, or may be a vector that has been subjected to appropriate processing on the vector. Examples of such processing include binning processing that converts continuous values ​​into discrete values, or normalization processing, which does not substantially change the magnitude relationship of each component in the vector before processing. These processings are as described above. Hereinafter, the vector (I FSC ,I p-FSC ,ISSC ,Iλ1,Iλ2,Iλ3) will be used as appropriate.

[0096] Next, the information processing device generates a representative vector consisting of some or all of the components of the feature vector. The representative vector preferably includes at least a component corresponding to the scattered light intensity and a component corresponding to the luminescence, among the components of the feature vector. In the above example, the representative vector is, for example, a vector (I FSC ,I p-FSC ,Iλ1,Iλ2,Iλ3).

[0097] Which of the components of the feature vector is used to generate the representative vector may be set by the user of the information processing device, or may be set by the information processing device based on a predetermined algorithm. The information processing device generates a representative vector for each extracellular particle using the set components of the feature vector. The representative vector may be the same vector as the feature vector.

[0098] Next, the information processing device determines the extracellular particles to be analyzed from the population of extracellular particles (S232). At this time, all of the extracellular particles in the population of extracellular particles may be used for the analysis, or a portion of a plurality of extracellular particles contained in the population of extracellular particles may be used for the analysis. The determination of the extracellular particles to be used for the analysis is performed, for example, based on the feature vector or the representative vector. The information processing device may not determine the extracellular particles to be used for the analysis, but may use all of the extracellular particles in the population of extracellular particles for the analysis.

[0099] The information processing device may, for example, exclude from the analysis extracellular particles for which all of the multiple components corresponding to the multiple emissions of the components of the representative vector are below the threshold. In the above example, this corresponds to excluding from the analysis extracellular particles for which Iλ1, Iλ2, and Iλ3 are all below the threshold. This corresponds to excluding from the analysis extracellular particles that are not labeled or stained, or that have a low degree of labeling or staining, from the group of extracellular particles. In this way, by excluding from the analysis extracellular particles that are not labeled or stained, or that have a low degree of labeling or staining, the analysis described below can be performed with higher accuracy.

[0100] Alternatively, the information processing device may use only extracellular particles for which a predetermined component is equal to or greater than a threshold value among a plurality of components corresponding to a plurality of emissions of the components of the representative vector, or may exclude such extracellular particles from the analysis. This aspect corresponds to adding to or excluding from the analysis extracellular particles having a high or low content of a predetermined component among a group of extracellular particles.

[0101] The above thresholds may be set by a user of the information processing device, or may be set by the information processing device based on a predetermined algorithm, similar to the pre-processing of data in methods for assisting in the diagnosis or prognosis prediction of a disease.

[0102] Next, in one embodiment, the information processing device creates a two-dimensional or three-dimensional scatter plot by mapping the extracellular particles analyzed above on a two-dimensional or three-dimensional space based on the similarity of their respective representative vectors (S233). This mapping may be performed, for example, by reducing the dimension of the representative vectors corresponding to each extracellular particle, or by machine learning. More specifically, techniques such as t-SNE (T-distributed Stochastic Neighbor Embedding), SNE, UMAP (Uniform Manifold Approximation and Projection), and principal component analysis (PCA) can be used. In the two-dimensional or three-dimensional scatter plot created, multiple extracellular particles are plotted based on the similarity of their representative vectors. That is, extracellular particles with similar representative vectors (corresponding to similar sizes and staining degrees) are plotted adjacent to each other, and extracellular particles with different representative vectors (corresponding to different sizes and staining degrees) are plotted apart.

[0103] Next, the information processing device labels each point corresponding to each extracellular particle in the scatter plot based on a plurality of components corresponding to at least a plurality of luminescences among the components of the representative vector. Examples of labels to be attached include the type of component or labeling substance from which luminescence was detected and the peak wavelength of the luminescence. For example, labeling may be performed with a component, labeling substance, or staining reagent corresponding to the strongest component among the components corresponding to the luminescence intensity in the feature vector or representative vector, or labeling may be performed using the following method. That is, for each labeling substance, it is determined whether each extracellular particle in the population of extracellular particles is labeled with the labeling substance, the target substance with the fewest number of labeled extracellular particles is designated as the first labeling substance, the target substance with the second fewest number of labeled extracellular particles is designated as the second labeling substance, the extracellular particles labeled with the first labeling substance are labeled with the first labeling substance, and the extracellular particles labeled with the second labeling substance are labeled with the second labeling substance, and this labeling is repeated the number of times equal to the number of labeling substances. In this labeling, the label to be attached to each extracellular particle may be set by the user of the information processing device, or may be set by the information processing device based on a predetermined algorithm.

[0104] In this manner, at least a portion of a plurality of extracellular particles contained in a population of extracellular particles is mapped, and a scatter diagram (first scatter diagram) labeled with a predetermined label can be created.

[0105] 5, in one embodiment, following step S232, the information processing device classifies the extracellular particles contained in the population of extracellular particles into two or more clusters based on the similarity of the representative vectors (S235), in addition to and / or instead of step S233. This embodiment differs from the embodiment in which steps S233 and S234 described in detail above are performed in that the population of extracellular particles is classified into two or more clusters based on the similarity of the representative vectors without creating a first scatter plot. As a clustering method, hierarchical clustering and non-hierarchical clustering can be used, and for example, a single-link method, a Ward method, a k-means method, a k-means++ method, etc. may be used.

[0106] The number of clusters to be generated is not particularly limited as long as it is two or more, and may be, for example, within the range of the number of detected light emissions ±5, the range of the number of detected light emissions ±4, the range of the number of detected light emissions ±3, the range of the number of detected light emissions ±2, or the range of the number of detected light emissions ±1, or the same number as the number of detected light emissions. Alternatively, if the number of components of the representative vector that correspond to light emissions is n, then clusters may be generated within the range of n±5, n±4, n±3, n±2, n±1, or n.

[0107] The information processing device may record cluster information about the type of cluster into which extracellular particles belonging to the generated clusters have been classified, in association with the feature vectors corresponding to the extracellular particles.

[0108] While an example of information processing in the method for analyzing extracellular particles of this embodiment has been described using Figure 5, it goes without saying that the information processing in the analysis method of this embodiment is not limited to the above example. For example, data processing such as feature engineering for machine learning may be performed before or during the above information processing. For example, binning processing may be performed on each feature vector obtained from a population of extracellular particles based on scattered light intensity and / or the intensity of luminescence derived from each labeling substance.

[0109] Furthermore, in the analysis method of this embodiment described with reference to FIG. 5, following step S234, the multiple extracellular particles mapped on the first scatter plot may be classified into two or more clusters based on the first scatter plot created by the information processing device. The classification into two or more clusters (also referred to as clustering) may be performed based on the labels of the first scatter plot. For example, clustering may be performed so that extracellular particles included in areas of the first scatter plot where the same labels are concentrated are included in the same cluster. Clustering may be performed by a user, or may be performed by an information processing device based on a predetermined algorithm. For example, algorithms that may be used include k-means, k-means++, support vector machine (SVM), kernel SVM, etc.

[0110] The number of clusters to be generated may be within the range described in step S235. Alternatively, if the number of labels in the first scatter plot is m, then clusters may be generated in the range of m±5, m±4, m±3, m±2, m±1, or m.

[0111] The information processing device may record cluster information about the type of cluster into which extracellular particles belonging to the generated clusters have been classified, in association with the feature vectors corresponding to the extracellular particles.

[0112] 5 may further include identifying characteristics of at least one of the clusters obtained by the clustering. The identification may include creating a second scatter plot by an information processing device and identifying characteristics of the cluster based on the first and second scatter plots. The creation of the second scatter plot and the identification of characteristics of the cluster based on the first and second scatter plots will be described in detail below.

[0113] The second scatter diagram is a scatter diagram in which at least a part of a plurality of extracellular particles contained in a group of extracellular particles is mapped in the same manner as the first scatter diagram, and one of a plurality of components corresponding to a plurality of luminescences of the components of the representative vector is visualized for each of the mapped plurality of extracellular particles. That is, the second scatter diagram differs from the first scatter diagram in that, instead of labeling based on a plurality of components corresponding to at least a plurality of luminescences of the components of the representative vector, information on the magnitude of the value of one of the plurality of components corresponding to a plurality of luminescences of the components of the representative vector, that is, the intensity of one of the plurality of luminescences is visualized and given.

[0114] In the second scatter plot, the visualized information may be visualized by a change in color or by the size of each point corresponding to the extracellular microparticles. For example, the second scatter plot is a heat map showing the degree of staining of the extracellular microparticles with a predetermined staining reagent.

[0115] The second scatter diagram is created by an information processing device. The information processing device creates the scatter diagram by mapping at least a portion of the multiple extracellular particles contained in the extracellular particle population in two or three dimensions, similar to the first scatter diagram. Next, the information processing device assigns information on the magnitude of one of the multiple components corresponding to the multiple emissions of the components of the representative vector to each point corresponding to each extracellular particle in the scatter diagram. A vector (I FSC ,I p-FSC In the above example where Iλ1, Iλ2, Iλ3 are obtained, the information processing device assigns information on the magnitude of Iλ1, Iλ2, Iλ3 to each point corresponding to each extracellular particle in the scatter diagram.

[0116] The information processing device may create a plurality of second scatter diagrams in which the types of light emissions visualized are different from each other. That is, when the representative vector includes a component corresponding to the first light emission and a component corresponding to the second light emission, the information processing device may create a second scatter diagram A in which the component corresponding to the first light emission is visualized, and a second scatter diagram B in which the component corresponding to the second light emission is visualized. A vector (I FSC ,I p-FSC In the above example where Iλ1, Iλ2, Iλ3 are obtained, the information processing device may create three types of second scatter plots in which information on the magnitude of Iλ1, Iλ2, and Iλ3 is assigned to each point on the scatter plot.

[0117] The identification of the characteristics of the cluster based on the first and second scatter plots may be performed by comparing the first and second scatter plots. The identification of the characteristics of the cluster may be performed by comparing the first scatter plot with a plurality of second scatter plots. For example, the vector (I FSC ,I p-FSC For example, when three types of second scatter plots are created in which the magnitude information of Iλ1, Iλ2, and Iλ3 is assigned to each point of the scatter plot, by referring to the three types of second scatter plots, it is possible to confirm the degree of staining by the three staining reagents with emission wavelengths of λ1, λ2, and λ3 in at least one cluster in the first scatter plot, and to identify the characteristics such as the content of each component of the extracellular microparticles.

[0118] (Extracellular particle separation) The analytical method of this embodiment can acquire analytical data including a combination of multiple feature vectors and cluster information, and by using such analytical data, it is possible to separate extracellular particles having specific properties from a population of extracellular particles. In such separation, specific extracellular particles may be separated from the population of extracellular particles analyzed by the analytical method of this embodiment, or specific extracellular particles may be separated from a population of extracellular particles different from the population of extracellular particles analyzed by the analytical method of this embodiment.

[0119] In such a separation method, a feature vector is acquired for the extracellular particles in the same manner as in the analysis method shown in Figure 5, and based on the feature vector and analysis data including a combination of multiple feature vectors and cluster information, it is determined whether the extracellular particles from which the feature vector has been acquired have predetermined characteristics, and extracellular particles determined to have the predetermined characteristics may be separated from the group of extracellular particles.

[0120] In this separation method, the population of extracellular particles from which the predetermined extracellular particles are separated may be the same as or different from the population of extracellular particles used to obtain the analytical data. For example, analytical data including a combination of a plurality of feature vectors and cluster information may be obtained by analyzing a first population of extracellular particles using the analytical method of this embodiment, and the predetermined extracellular particles may be separated from the second population of extracellular particles based on the analytical data. Alternatively, analytical data including a combination of a plurality of feature vectors and cluster information may be obtained by analyzing a first population of extracellular particles using the analytical method of this embodiment, and the predetermined extracellular particles may be separated from the first population of extracellular particles based on the analytical data.

[0121] In this separation method, the characteristics of the extracellular particles to be separated may be set by the user. For example, the user may instruct the information processing device to separate extracellular particles belonging to one of the clusters of extracellular particles discovered by the analysis method of this embodiment. In this case, the information processing device analyzes the characteristics of the extracellular particles belonging to the specified cluster based on the analysis data. For example, the information processing device references the feature vectors or representative vectors of the extracellular particles belonging to the specified cluster and determines the ranges of each component of the feature vectors or representative vectors that indicate the need to separate the extracellular particles. Alternatively, the information processing device may generate a predictive model that predicts whether a given extracellular particle belongs to a specified cluster. To generate the predictive model, training data may be used, including the feature vectors or representative vectors of the extracellular particles belonging to the specified cluster and cluster information. The predictive model may be a regression model or a model generated by machine learning. Methods such as convolutional neural networks (CNNs), decision trees, random forests, naive Bayes methods, and support vector machines may be used to generate the model.

[0122] More specifically, the extracellular particles to be separated may be determined as follows: First, the user instructs the information processing device to separate extracellular particles belonging to one of the clusters of extracellular particles found by the analysis method of this embodiment. The information processing device identifies labeled or stained components of the extracellular particles in the specified cluster based on analysis data including a combination of multiple feature vectors and cluster information obtained by analyzing a group of extracellular particles using the analysis method of this embodiment. The presence or absence of labeling or staining may be determined based on the intensity of luminescence derived from the labeling substance or staining reagent or the components of the corresponding feature vector or representative vector. For example, extracellular particles whose luminescence intensity or the components of the corresponding feature vector or representative vector are equal to or greater than a threshold may be determined to be labeled or stained. Next, the information processing device acquires feature vectors related to the extracellular particles in the group of extracellular particles to be separated, and identifies labeled or stained components of the extracellular particles based on the feature vectors. The information processing device determines that the extracellular particles are to be separated when the labeled or stained constituents of the extracellular particles are the same as the labeled or stained constituents of the extracellular particles of the designated cluster.

[0123] The method for separating extracellular particles determined to have predetermined properties from a group of extracellular particles is not particularly limited as long as it is a method that can separate extracellular particles at the single particle level, and examples thereof include a method using a flow cytometer. For example, a method may be used in which droplets containing extracellular particles are formed, the droplets containing the extracellular particles to be separated are charged, and only the droplets containing the extracellular particles to be separated are electrically separated.

[0124] In this way, a plurality of extracellular particles having predetermined properties can be separated to obtain a new population of extracellular particles. The obtained population of extracellular particles may be analyzed again by the analytical method of this embodiment to perform further cluster analysis, or may be subjected to other analyses such as PCR.

[0125] [Information acquisition / analysis device] Fig. 6 is a diagram showing an example of the configuration of an information acquisition and analysis device for carrying out the method for assisting in the diagnosis or prognosis prediction of a disease and / or the analysis method of this embodiment. As shown in Fig. 6, the device 1 of this embodiment includes a detection unit 2 that detects, for each extracellular particle in a population of extracellular particles, scattered light from irradiated light and multiple emissions with different peak wavelengths that are generated from the constituent components of the extracellular particles or from labeling substances bound to the constituent components due to the irradiated light; and an information processing unit 3 that performs information processing based on the detected scattered light and multiple emissions. The device 1 of this embodiment also includes an input unit 4 that allows a user to give instructions and input data to the information processing unit and the detection unit, and an output unit 5 that outputs analysis results. Each component will be described below, and the above description will be omitted as appropriate.

[0126] The detection unit 2 detects scattered light from the irradiated light and multiple luminescence emissions with different peak wavelengths that are generated from the constituent components of the extracellular particles or from labeling substances bound to the constituent components due to the irradiated light for each extracellular particle in the population of extracellular particles. In Figure 6, the detection unit 2 includes a sample introduction means for introducing a sample to be analyzed, a flow means for flowing the introduced sample through a flow path to create a state in which the extracellular particles flow continuously in a substantially straight line, a light irradiation means for irradiating light onto the extracellular particles continuously flowing in a substantially straight line by the flow means, and a light detection means for detecting scattered light and multiple luminescence emissions with different peak wavelengths from the irradiated extracellular particles. The detection unit 2 may have any configuration as long as it can detect scattered light and multiple luminescence emissions with different peak wavelengths for each extracellular particle in the population of extracellular particles as described above, and for example, a commercially available flow cytometer may be used as the detection unit.

[0127] The flow means flows the introduced sample through the flow channel, creating a state in which the extracellular particles flow continuously in a substantially single file. The flow means may include a flow cell having a sample inlet and a sample outlet, and a pressure applying means such as a pump that introduces the sample into the flow cell. The pressure applied by the pressure applying means is adjusted within a range that generates a laminar flow in the flow cell.

[0128] The light irradiating means irradiates light having an excitation wavelength of a component contained in the extracellular microparticle or a labeling substance bound to the component. The irradiated light may be white light or monochromatic light. When the population of extracellular microparticles is labeled with multiple types of labeling substances that emit excitation light of different wavelengths, the detection unit 2 may be provided with multiple light irradiating means that irradiate multiple types of monochromatic light corresponding to the wavelengths of the excitation light.

[0129] The light detection means detects scattered light and multiple luminescence beams with different peak wavelengths generated from the extracellular microparticles for each particle. The light detection means may include a spectroscope such as a diffraction grating, bandpass filter, and notch filter, a light-collecting element such as a lens, a detector such as a photomultiplier tube and photodiode, and a polarizing element for detecting polarized light components. More specifically, the light detection means may include a light-collecting element; a spectroscope and detector for detecting scattered light; a spectroscope, polarizing element, and detector for detecting polarized scattered light; and multiple spectroscopes and detectors for detecting multiple luminescence beams with different peak wavelengths.

[0130] The input unit 4 accepts input to the device 1. The input unit 4 includes, for example, a keyboard, a mouse, a microphone, and / or a touch panel for accepting input from a user. The device 1 causes the detection unit 2, the information processing unit 3, and the output unit 5 to perform their respective functions based on the information input from the input unit 4. The input unit 4 may also accept analytical data obtained by analyzing a population of extracellular particles using the analytical method described in detail above. The input unit 4 may also accept disease information regarding the disease state of a subject.

[0131] The output unit 5 displays the analysis results of the device 1. The output unit 5 includes, for example, a liquid crystal display or an organic EL display.

[0132] Fig. 7 is a block diagram showing an example of the functional configuration of the information processing unit 3. In Fig. 7, the information processing unit 3 includes a communication means for communicating with the detection unit 2, the input unit 4, and the output unit 5, an extracellular particle information acquisition means for acquiring extracellular particle information including scattered light information and luminescence information from the detection unit 2, a disease information acquisition means for acquiring disease information related to the state of the disease in the subject, a processing target determination means for determining a target to be used in the analysis means and / or auxiliary information creation means from the acquired extracellular particle information, a data processing means for processing data obtained from the detection unit 2 and the extracellular particle information acquired by the extracellular particle information acquisition means, an analysis means for performing the analysis method of this embodiment, and an auxiliary information creation means for performing the method for assisting in the diagnosis or prognosis prediction of a disease of this embodiment.

[0133] The information processing unit 3 may be integrated with the detection unit 2, the input unit 4, and / or the output unit 5 to form the device 1, or may be connected to the detection unit 2, the input unit 4, and / or the output unit 5 via a wired or wireless connection to form the device 1. The information processing unit 3 communicates with other components of the device 1 via a communication means. For example, the information processing unit 3 may receive detection data from the detection unit via communication from the input unit, and transmit the processing results to the output unit to display them on the output unit. The communication means is not particularly limited and may be processing within the device, or wireless or wired communication.

[0134] The data processing means processes the detection data received from the detection unit 2 for subsequent analysis. For example, it may correspond each extracellular particle to the detected scattered light intensity and luminescence intensity based on the detection data. The data processing means may also perform luminescence spillover correction. The luminescence spillover correction is a correction that reduces the influence of multiple luminescence emissions with different peak wavelengths on each other's detected intensity. For example, when a first luminescence exhibits broad luminescence and its peak wavelength overlaps with the peak wavelength of a second luminescence with a different peak wavelength, the correction corresponds to subtracting the intensity attributable to the first luminescence from the intensity of the second luminescence. The data processing means may also perform the above-mentioned data preprocessing.

[0135] The disease information acquiring means acquires disease information of each subject input using the input unit 4. The disease information may be stored in association with extracellular particle information for each subject.

[0136] The extracellular particle information acquiring means acquires extracellular particle information including scattered light information and luminescence information from the detection unit 2. At this time, the extracellular particle information may be acquired after preprocessing of the data via the data processing means. The extracellular particle information acquiring means may also include a feature vector generating means and a representative vector generating means, which may generate the feature vectors and representative vectors. The feature vector generating means and the representative vector generating means may generate the feature vectors and representative vectors, respectively, in response to instructions from the input unit 4 or based on a predetermined algorithm, for example.

[0137] The extracellular particle information is acquired for each extracellular particle in the population of extracellular particles for each subject. The processing target determination means determines which of the acquired extracellular particle information for each extracellular particle is to be used for processing in the analysis means and / or auxiliary information creation means described below. For example, as described above, the processing target determination means may exclude from the population of extracellular particles extracellular particles that are not labeled or stained, or that are only lightly labeled or stained.

[0138] The analysis means performs the analysis method of the present embodiment described above for analyzing a group of extracellular particles collected from the lungs or exhaled breath of the subject based on the acquired extracellular particle information. The analysis means may include mapping means for two-dimensionally or three-dimensionally mapping a plurality of extracellular particles determined as mapping targets based on the similarity of the representative vectors, labeling means for labeling each of the mapped plurality of extracellular particles, component visualization means for visualizing one of a plurality of components corresponding to a plurality of emissions of the components of the representative vector in each of the mapped plurality of extracellular particles, and cluster generation means for classifying the analysis targets into two or more clusters.

[0139] The mapping means may perform mapping by machine learning based on the representative vectors corresponding to each extracellular particle generated by the representative vector generating means, for example.

[0140] The label that the labeling means attaches to each extracellular particle may be set by an instruction from the input unit 4 or based on a predetermined algorithm.

[0141] The component visualization means may visualize the luminescence intensity by color, or may visualize by the size of each point corresponding to the extracellular particle. The components of the representative vector visualized by the component visualization means may be set by instructions from the input unit 4.

[0142] The cluster generating means may assign the same cluster information to extracellular particles that fall within a predetermined range in the first scatter plot. The range to which the same cluster information is assigned may be set by an instruction from the input unit 4 or based on a predetermined algorithm.

[0143] The auxiliary information creating means carries out the method of this embodiment described above to assist in the diagnosis or prognosis prediction of a disease based on the acquired extracellular particle information and disease information. The auxiliary information creating means may include at least one of a diagnostic auxiliary information creating means for creating information to assist in diagnosis, a prognosis prediction auxiliary information creating means for creating information to assist in prognosis prediction, a trained diagnostic model, and a trained prognosis prediction model.

[0144] The diagnostic auxiliary information creation means may determine what diagnostic auxiliary information to obtain from the disease information and extracellular particle information based on instructions from the input unit 4 or on a predetermined algorithm, and obtain the diagnostic auxiliary information from the disease information and extracellular particle information based on this.

[0145] The prognosis prediction auxiliary information creation means may determine what kind of prognosis prediction auxiliary information to acquire from the disease information and extracellular microparticle information based on instructions from the input unit 4 or based on a predetermined algorithm, and acquire the prognosis prediction auxiliary information from the disease information and extracellular microparticle information based on this.

[0146] The diagnostic model and prognosis prediction model may be pre-stored trained models, trained models created by the diagnostic auxiliary information creation means and the prognosis prediction auxiliary information creation means, or trained models input from the input unit 4.

[0147] Fig. 8 is a block diagram showing an example of the physical configuration of the information processing unit 3. In Fig. 8, the information processing unit 3 has a RAM (random access memory) 31, a ROM (read only memory) 32, a storage 33, a CPU (central processing unit) 34, a receiving means 35, and a transmitting means 36, as well as a system bus 37 connecting these.

[0148] The RAM 31 is a rewritable memory that serves as a main memory. The RAM 31 may be configured, for example, with a semiconductor memory device, and stores various data and programs such as applications executed by the CPU .

[0149] The ROM 32 is a memory that can only read data and may be configured with, for example, a semiconductor memory element. The ROM 32 stores programs such as firmware and data.

[0150] The storage 33 is a rewritable memory that serves as an auxiliary memory. The storage 33 may be configured, for example, as a semiconductor memory element, an optical disk, a hard disk drive (HDD), or a magnetic tape, and stores programs and various data.

[0151] The CPU 34 is a control unit that controls the execution of programs stored in the RAM 31 and / or ROM 32, and performs calculations and processing of data. The information processing unit 3 realizes the functions related to the information processing according to the present embodiment described above under the control of the CPU 34. The CPU 34 executes information processing based on information and instructions received from the receiving means 35, and transmits the processing results via the transmitting means 36 or stores them in various storage devices such as the RAM 31 and the storage 33.

[0152] The receiving means 35 is a means for receiving information and instructions from the detection unit 2 and the input unit 4 into the information processing unit 3, and the transmitting means 36 is a means for transmitting information and processing results from the information processing unit 3 to the detection unit 2 and the output unit 5. The receiving means 35 and the transmitting means 36 may be wireless or wired communications.

[0153] In the information processing unit 3, the CPU 34 executes the program of this embodiment, thereby realizing the various functions described with reference to Fig. 7. Note that these physical configurations are merely examples, and they do not necessarily have to be independent configurations. For example, the information processing unit 3 may be provided with an LSI (Large-Scale Integration) in which the CPU 34, RAM 31, ROM 32, and / or storage 33 are integrated.

[0154] Although the device 1 of this embodiment has been described above with reference to Figures 6 to 8, the device 1 of this embodiment is not limited to this and various modifications are possible without departing from the spirit of the present invention. For example, the device 1 may be provided with a functional or physical configuration not shown in Figures 6 to 8, or the functional or physical configuration shown in Figures 6 to 8 may be omitted.

[0155] [Note] The present invention includes the following embodiments. [1] 1. A method for aiding in the diagnosis or prognosis of a disease, comprising: From each subject in the population of subjects, (A) extracellular particle information on a population of extracellular particles collected from the lungs or exhaled breath of the subject, the extracellular particle information including, for each extracellular particle of the population of extracellular particles, scattered light information on scattered light of irradiated light and luminescence information on a plurality of luminescences with different peak wavelengths that are generated from components of the extracellular particles or labeled substances bound to the components due to the irradiated light; and (B) disease information regarding the disease status of the subject; and Obtaining information for assisting in the diagnosis or prognosis prediction of the disease associated with extracellular particles collected from the lungs or exhaled breath by machine learning based on a dataset including the extracellular particle information and the disease information; A method comprising: [2] the population of subjects comprises a first population of subjects affected with the disease and a second population of subjects not affected with the disease; The disease information includes information on whether the subject is suffering from the disease. [1] The method described in [1]. [3] and acquiring characteristics of the extracellular microparticles characteristic of the first population through the machine learning. [2] The method described in [2]. [4] the machine learning identifies markers for the disease; [2] The method described in [2]. [5] creating a diagnostic model for the disease through the machine learning; The method according to [2] or [3]. [6] the population of subjects comprises a population of subjects suffering from a disease; the disease information includes a prognosis status of the subject; [1] The method described in [1]. [7] A patient prognosis prediction model is created by the machine learning. [6] The method described in [6]. [8] [5] or [7], A model for diagnosing said disease or for predicting the prognosis of a patient. [9] the disease is a pulmonary disease; The model described in [8].

[10] A diagnostic or prognostic model for non-squamous lung cancer, comprising: receiving extracellular particle information relating to a population of extracellular particles collected from the lungs or exhaled breath of a subject, the extracellular particle information including, for each extracellular particle of the population of extracellular particles, scattered light information relating to scattered light of irradiated light and luminescence information relating to a plurality of luminescences with different peak wavelengths that are caused by the irradiated light from components of the extracellular particles or labeled substances bound to the components; making a diagnosis or prognosis prediction based on the concentration of at least one of EpCAM-positive extracellular microparticles and podoplanin-positive extracellular microparticles in said population of extracellular microparticles; Model.

[11] 1. A method for analyzing a sample collected from the lungs or exhaled air and containing a population of extracellular particulates labeled with a plurality of labeling substances, comprising: acquiring extracellular particle information for each extracellular particle contained in the group of extracellular particles, the information including scattered light information regarding scattered light of the irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are generated from components of the extracellular particles or labeling substances bound to the components due to the irradiated light; analyzing the population of extracellular particles by machine learning based on the extracellular particle information; A method comprising:

[12] analyzing the population of extracellular particulates generating a feature vector for each extracellular particle included in the population of extracellular particles based on the acquired extracellular particle information; mapping at least some of the extracellular particles contained in the group of extracellular particles in two or three dimensions based on the similarity of a representative vector consisting of some or all of the components of the feature vector, and labeling each of the mapped extracellular particles based on at least a plurality of components of the representative vector that correspond to the luminescence information, thereby creating a first scatter diagram; The method according to

[11] , comprising:

[13] analyzing the population of extracellular particulates generating a feature vector for each extracellular particle included in the population of extracellular particles based on the acquired extracellular particle information; classifying the extracellular particles contained in the group of extracellular particles into two or more clusters based on the similarity of a representative vector consisting of some or all of the components of the feature vector; The method according to

[11] or

[12] , comprising:

[14] 1. A program for analyzing a sample collected from the lungs or exhaled breath of a subject, the sample including a population of extracellular particles labeled with a plurality of labeling substances, the program comprising: Computer, an extracellular particle information acquiring means for acquiring extracellular particle information for each extracellular particle contained in the group of extracellular particles, the extracellular particle information including scattered light information regarding scattered light of irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are generated from components of the extracellular particles or labeling substances bound to the components due to the irradiated light; an analysis means for analyzing the population of extracellular particles by machine learning based on the extracellular particle information; A program that functions as a

[15] Computers, and more a disease information acquiring means for acquiring disease information relating to the state of a disease in the subject; an auxiliary information generating means for generating information for assisting in the diagnosis or prognosis prediction of the disease associated with the extracellular particles collected from the lungs or exhaled breath by machine learning based on a dataset including the extracellular particle information and the disease information;

[14] The program according to

[14] ,

[16] Computers, and more The obtained extracellular particle information is input into a model created by the method described in [5] or [7] to produce a diagnostic model for diagnosing a disease or a prognosis prediction model for predicting the prognosis of the subject. The program according to

[14] or

[15] , which functions as

[17] A computer-readable recording medium on which any one of the programs described in

[14] to

[16] is recorded. Here, the recording medium may be a non-transitory tangible medium such as a CD-ROM, an SD card, or a USB memory. [Example]

[0156] The present invention will be described in more detail below using examples and comparative examples, but the present invention is not limited to the following examples.

[0157] 1. Method [BD Influx cell sorter setup] A BD Influx cell sorter (BD Biosciences) equipped with 355, 405, 488, 561, and 640 nm lasers was installed in a Class II Type AII biosafety cabinet (Baker). To collect forward scatter signals, an optical system consisting of a high-NA, long-working-distance 20x objective lens, a 0.7 mm diameter pinhole, and two photomultiplier tubes (PMTs) detecting light with different polarization directions (FSC-par and FSC-per) was used. When the FSC sensitivity was adjusted to detect 100 nm particles, 500 nm particles showed a saturated signal in FSC, while polarization FSC (FSC-perp) could be used to analyze larger particles (>200 nm). Therefore, FSC and FSC-perp were used to analyze a wide range of extracellular particle sizes. Since it has been reported that an 8 mm obscuration bar is superior to the conventional 5 mm bar in reducing noise signals in FSC (Arkesteijn, GJA et al. Improved Flow Cytometric Light Scatter Detection of Submicron-Sized Particles by Reduction of Optical Background Signals. Cytometry A 97, 610-619 (2020).), we fabricated an 8 mm aluminum bar on a 2 mm bar.

[0158] The nozzle size (70, 86, 100, and 140 μm) did not significantly affect the resolution of extracellular particle analysis, but increasing the sheath fluid flow rate reduced the signal from the particles. Therefore, a 70 μm nozzle was selected, and the sheath pressure was adjusted to 24.0 psi. Laser alignment was optimized with ultra rainbow fluorescent particles (Spherotech) according to the user guide. All signals were collected by a PMT detector and displayed on a logarithmic scale (10 0 ~10 4) and displayed as a height signal. To determine the optimal threshold trigger for extracellular particle analysis, the FSC signal voltage was adjusted to 33 and 100 nm fluorescent beads (Invitrogen) were measured. An FSC threshold of 0.58 to 0.6 was adopted. The FSC-perp voltage was set to 33 (for comparison with FSC) or 14 (for analysis of large particles and cells). For extracellular particle analysis, the signal intensity above noise was adjusted to 10 1 The fluorescence voltage was adjusted to be less than 100 kJ / s. Unless otherwise noted, the sample pressure was set between 24.9 and 25.2 psi, and the detection rate was not to exceed 30,000 per second.

[0159] [Fluorescent bead sorting] Beads of 100 nm (Invitrogen) and 200, 500, and 1000 nm (Polysciences) were mixed in PBS and analyzed on a BD Influx cell sorter. In the sorting setup, droplet formation not only impaired the SSC signal but also increased the number of noise signals at the FSC threshold trigger, which is thought to be due to a change in the angle of incidence of the laser beam on the sheath liquid surface due to droplet formation. We found that this effect could be reduced by lowering the droplet nozzle position and breakoff position. Based on these preliminary findings, we first adjusted the drop frequency between 60.2 and 60.8 to obtain the highest breakoff position. Next, the piezo amplitude was set to the lowest value that formed a stable test flow (typically 3.0–4.5). Charging timing was optimized using AccuDrop beads (BD Biosciences). The optimal charging timing changed with increasing noise. Target droplets failed to be collected at the optimal charging timing with an FSC threshold of 0.6. Therefore, we found that the charging timing should be optimized under an increased FSC threshold trigger (approximately 3.0) to eliminate noise signals. After adjusting the charging timing, the FSC threshold trigger was returned to 0.58-0.6. Target particles were then sorted in bidirectional sorting mode and collected into uncoated 1.5 or 5 mL tubes.

[0160] A mixture of 100, 200, 500, and 1000 nm beads was also measured using the original optical system on a BD Aria III cell sorter and a BD LSRFortessa. Detection of 100-500 nm particles with the BD Aria III or 100-200 nm particles with the BD LSRFortessa was difficult due to saturated noise signals triggered by FSC threshold. The FSC voltage and threshold were adjusted to achieve noise signals below 5000 readings per second. Detection of fluorescent beads using a CytoFLEX LX (Beckman Coulter) was performed using a violet SSC threshold according to the small particle analysis user guide. Sorting of 1000 nm particles with the BD Aria III was performed in purity mode. Reanalysis of sorted fractions was performed on a BD Influx. Purity of the sorted samples was calculated as the ratio of each population to the sum of the 100, 200, 500, and 1000 nm populations.

[0161] [Sample preparation] C57BL / 6J mice (6-8 weeks old) were purchased from CLEA, Japan. Mouse serum was collected by centrifugation of blood at 2,000 × g for 10 min and kept at -80 °C until use. For analysis of mouse BALF, mice were anesthetized and spun down with 0.1–10 LD in 30 μL of PBS. 50 The mice were intranasally infected with the A / Puerto Rico / 8 / 1934 influenza (H1N1) virus.

[0162] Human bronchoalveolar lavage fluid (BALF) was obtained from surplus specimens collected for diagnostic purposes. Specifically, a flexible bronchoscope was inserted into the bronchioles of the middle lobe or lingule. 50 mL of sterile saline was then added and manually aspirated three times. The BALF was filtered through sterile gauze and centrifuged to remove cells. The supernatant was stored at -80°C until use. Total cell counts were determined using a hemocytometer.

[0163] Human exhaled breath coagulates (EBCs) were collected using R-tubes (Respiratory Research Inc.) from individuals with lung cancer, nontuberculous mycobacterial infection, or healthy individuals. Donors were instructed to inhale through their nose and exhale through their mouths at a normal breathing rhythm for 9–10 minutes. Participants did not wear nose clips. Samples were then stored at −80°C until use.

[0164] [Feature Engineering of Flow Cytometry Data for Machine Learning] To utilize the data from individual BALF flow cytometry analyses for machine learning, we generated features from the raw flow cytometry data by binning. Specifically, each particle detected by flow cytometry contains multidimensional information consisting of scattered light and fluorescent signals. Therefore, to obtain features of extracellular particle concentrations in BALF while retaining as much information as possible, we divided particles into seven size groups (three and four, respectively) based on FSC and vertically polarized FSC signals, as shown in Figure 11C. Furthermore, each particle size was further divided into two bins for each fluorescence intensity. By repeating this process for all fluorescent signals, particles in BALF and EBC were classified into 3584 and 256 bins, respectively. These procedures were performed using FlowJo software. After removing blank bins, count data from 1490 and 174 bins were used as features for the BALF and EBC samples, respectively.

[0165] [Machine Learning Analysis] The machine learning framework was built in Python (version 3.10) using the following libraries: scikit-learn (version 1.0.2), numpy (version 1.21.5), and pandas (version 1.4.2). Data were normalized using StandardScaler or RobustScaler. PCA was performed using a scikit-learn pipeline. Prognostic prediction models were generated using LightGBM with normalized training data split by LOOCV. Diagnostic models were generated using linear SVC with normalized training data split by LOOCV. When clinical data for patients who provided exhaled breath coagulation (EBC) were missing, the mean value of available participants was used as input. Feature selection was performed using SelectKBest.

[0166] [Statistical analysis] Experiments were performed independently at least twice. Experimental results are shown as mean ± SEM. Statistical significance of differences between groups was determined by unpaired Student's t-test or one-way ANOVA with Tukey's multiple comparison test, log-rank test, or two-tailed Pearson's correlation. Significant differences were annotated with asterisks: P<0.05: *, P<0.01: **, P<0.001: ***.

[0167] 2. Detection and sorting of extracellular microparticles using dual FSC signals A major challenge in analyzing extracellular particles using flow cytometry is refining the threshold trigger to eliminate noise signals while maintaining the ability to detect nanoparticles. Fluorescence-based thresholding effectively separates fluorescently labeled particles from noise signals, but it excludes almost all non-labeled particles, making it unsuitable for comprehensive analysis (Figure 9A). New analyzers, such as nanoFCM and CytoFLEX, address this limitation by employing side scatter (SSC)-based thresholding, but building an in-sorter analysis system remains challenging. In particular, the size range of extracellular particles ranges from less than 100 nm to more than 1 μm, requiring a wide range of size resolution for comprehensive characterization.

[0168] To achieve high-parametric analysis and sorting of a wide range of extracellular particles, we used a BSL-2-compliant, high-throughput cell sorter, the BD Influx, equipped with five lasers (Figure 9B). Following previous reports (Arkesteijn, GJA et al. Improved Flow Cytometric Light Scatter Detection of Submicron-Sized Particles by Reduction of Optical Background Signals. Cytometry A 97, 610-619 (2020).; Nolte-'t Hoen, EN et al. Quantitative and qualitative flow cytometric analysis of nanosized cell-derived membrane vesicles. Nanomedicine 8, 712-720 (2012).), we employed an 8-mm obscuration bar to detect 100-nm fluorescent beads. We also used a polarization unit to cover a wide size range (Figure 9B). Because FSC thresholding showed better results in detecting 100 nm beads than polarized FSC (FSC-perp), small (100-200 nm) and large (>200 nm) particles were analyzed by normal FSC and polarized FSC, respectively (Figure 9C).

[0169] Next, we performed particle sorting tests using the BD Influx cell sorter. FSC thresholding with optimized optics enabled separation of fluorescent beads of different sizes with high purity (>99%) (Figure 9C). The BD Aria III, a conventional cell sorter equipped with a photodiode FSC, had difficulty detecting and sorting beads smaller than 1000 nm due to saturation of the noise signal (Figure 9C). When reanalyzed using the BD Influx cell sorter, the 1000 nm particle fraction sorted by the BD Aria III was significantly contaminated with other beads. This is likely due to the inability of the unoptimized BD Aria III to exclude droplets containing both 1000 nm and 100-500 nm particles (Figure 9C). These results demonstrate that FSC thresholding with optimized optics enables detection and sorting of particles across a wide size range.

[0170] 3. High-resolution characterization of extracellular microparticles based on multicolor staining To evaluate the potential of extracellular particle staining to identify heterogeneous extracellular particle populations, we performed eight-color staining of bronchoalveolar lavage fluid (BALF) containing extracellular particles by targeting nucleic acids, lipid membranes, glycans, and proteins without a washing step (Figure 10A). BALF was obtained from mice infected with influenza virus. This complex sample contains a wide range of extracellular particles, including virus particles and host-derived extracellular vesicles, and is suitable for use in the BSL-2 laboratory at BD Influx.

[0171] To detect membranous particles, we used CellMask deep red, a lipid membrane intercalator with negligible aggregate formation, and phospholipid-binding annexin V. SYBR gold and Hoechst 33258 were used as fluorescent probes for nucleic acids with different binding modes (Bis-intercalator and minor groove binder). To stain glycans, we used wheat germ agglutinin (WGA), which binds to sialic acid and N-acetylglucosamine commonly expressed on the surface of mammalian cells, and peanut agglutinin (PNA), a lectin that binds to terminal β-galactose exposed after neuraminidase cleavage. Proteins were stained with anti-hemagglutinin (HA) antibodies and amine-reactive dyes to detect virus particles and large aggregates, respectively.

[0172] Specifically, BALF was centrifuged at 400 × g for 5 minutes to remove cells, and the supernatant was collected for subsequent staining. When analyzing both extracellular particles and cells, BAL cells were concentrated 30-fold by centrifugation at 400 × g for 5 minutes. The sample was then diluted 5-fold with Annexin V staining buffer. Because the presence of other staining reagents prevented CellMask deep red from staining membranous particles, CellMask deep red was first mixed with the sample prior to the other staining reagents. The other staining reagents were then added to the following final concentrations: CellMask deep red (1 / 10,000), SYBR gold (1 / 5,000), Hoechst 33258 (1 / 100), Live / Dead stain Near-IR (Invitrogen, 1 / 1,000), BV510 Annexin V (1 / 20), CF405M WGA (1 / 200), CF568 PNA (1 / 20), RPE-CF647T anti-HA antibody (1 / 100), PE-CF594 anti-siglec-F antibody (BD Biosciences, 1 / 100), PE-Cy7 anti-CD11c antibody (Biolegend, 1 / 100), and BV785 anti-Ly6C antibody (Biolegend, 1 / 100). After 20 minutes of incubation, samples were diluted with Annexin V staining buffer and measured on a BD Influx cell sorter.

[0173] To visualize the heterogeneity of extracellular particles in BALF, we performed t-SNE (t-distributed stochastic neighbor embedding) analysis based on fluorescence and FSC signals, revealing that probe-stained particles appeared as a separate population from unstained particles (Figure 10D). Because unstained particles, primarily consisting of noise signals, appeared as the major population, reducing the resolution of the stained population, we next performed t-SNE analysis by setting an "or" gate to include only particles stained with at least one fluorescent probe (Figure 10E). These data processing techniques improved the resolution of AI-assisted extracellular particle mapping, revealing distinct populations for virus particles, glycan- or nucleic acid-containing particles, and other membranous particles (Figure 10F).

[0174] Data were analyzed using FlowJo (BD Biosciences, v10.7, v10.8, or v10.9) according to the following procedure. First, fluorescence signal compensation was performed using single-stained controls. The compensation value was determined as the minimum percentage that avoided detection of stained particles in other channels. Next, all particles, or stained particles extracted by applying an "or" gate, were analyzed by t-SNE (opt-SNE, iterations: 1000, perplexity: 30) or UMAP. All fluorescence and scattering parameters (FSC and vertically polarized FSC) for staining extracellular particles were included in the t-SNE analysis. When analyzing cells with t-SNE, cell surface markers were included only as parameters. The results of the t-SNE analysis were visualized as dot plots or counter plots with the colors indicated in the order of priority shown in each figure.

[0175] The importance of eliminating the wash step for complete characterization of extracellular particles in biological samples was evaluated compared with performing the wash step after staining. Specifically, stained extracellular particles were purified by ultracentrifugation (120,000 × g, 1 hour), qEV / 70nm with an Automatic Fraction Collector, or a 10 kDa Amicon Ultra centrifugal unit (10,000 × g, 10 minutes, Millipore). Similar analyses using size-exclusion chromatography, centrifugation, and filtration washes did not detect any nucleic acid-positive populations, suggesting that elimination of the wash step is preferable for complete characterization of extracellular particles.

[0176] Next, we further validated the quantitative performance of this analytical method by evaluating bronchoalveolar lavage fluid (BALF) samples from virus-infected mice with different disease severity levels. In this experiment, we stained not only extracellular particles but also cells by targeting surface markers with antibodies and analyzed them using the gating strategy described above and subsequent machine learning. Results showed that inflammatory cells, including neutrophils and monocytes, increased in response to influenza virus dose (Figures 10G and 10H). Analysis of stained extracellular particles showed that the abundance of each population correlated with disease severity (Figures 10G and 10H).

[0177] 4. Prognostic prediction of lung disease using multiparametric analysis of extracellular particles in human bronchoalveolar lavage fluid (BALF) To investigate whether the analytical methods demonstrated in sections 2 and 3 above could contribute to clinical evaluation, we performed multiparametric analysis on BALF samples from 132 patients with pulmonary disease using a BSL-2-compatible system (Figure 11A). A nine-color panel targeting several protein markers was constructed. Specifically, CellMask deep red (1 / 10,000), SYBR gold (1 / 5,000), Hoechst 33258 (1 / 100), CF405M WGA (1 / 2,000), LiveDead Near-IR (1 / 1,000), BV510 Annexin V (1 / 20), PE-Dazzle 594 CD9, PE anti-HLA-DR, DP, and DQ antibodies, and BV605 anti-HLA-A, B, and C antibodies (1 / 200, BioLegend) were used. The staining procedure was similar to that for mouse BALF.

[0178] Through multicolor staining and subsequent data processing, we found that human BALF contained multiple extracellular particle populations unique to each patient (Figure 11B). To analyze each patient's flow cytometry data using machine learning while retaining 10 dimensions (9 fluorescence parameters and FSC), we divided the stained extracellular particles into two parts based on each fluorescence signal and seven parts based on FSC signal, generating 3,584 bins of data (Figure 11C). By excluding blank bins, a total of 1,490 bins were used for machine learning. Unsupervised principal component analysis (PCA) suggested that the characteristics of extracellular particles in some patients differed from those in most patients (Figure 11D). Notably, these outlier patients exhibited significantly worse prognosis despite similar lung capacities (Figure 11D).

[0179] We then generated a prognostic model using data from 46 patients whose prognosis could be monitored. The model, created and validated using LightGBM and leave-one-out cross-validation (LOOCV), showed that the extracellular particle dataset outperformed the clinical dataset in predicting disease prognosis (Figure 11E).

[0180] 5. Comprehensive analysis of extracellular particles in patient breath condensates for non-invasive diagnosis of non-squamous cell lung cancer Finally, we demonstrated the feasibility of non-invasive lung cancer diagnosis using this embodiment. In this example, we demonstrated that squamous and non-squamous lung cancers, which require different treatment strategies, can be distinguished by analyzing extracellular particles in exhaled breath coagulation (EBC), a non-invasive liquid sample ( FIG. 12A ).

[0181] A six-color panel targeting several protein markers was constructed using CellMask deep red (1 / 10,000), SYBR gold (1 / 5,000), Hoechst33258 (1 / 100), CF405MWGA (1 / 2,000), LiveDead Near-IR (1 / 1,000), and BV510 Annexin V (1 / 40).

[0182] Multiparametric analysis of EBC samples revealed the presence of multiple particle populations similar to those observed in BALF (Figure 12B). PCA was performed using binned data of extracellular particle information from EBC samples from 14 lung cancer patients (squamous, n = 4; non-squamous, n = 10) and 10 control subjects (healthy, n = 9; nontuberculous mycobacteria, n = 1). The results revealed two non-squamous patients as outliers (Figure 12C, upper panel). Further analysis showed that the extracellular particle information of most non-squamous patients differed from squamous patients and healthy subjects in the remaining 22 subjects (Figure 12C, lower panel). While the EBC samples from the two non-squamous patients were indistinguishable from squamous patients and healthy subjects, one of them was in very poor condition with impaired respiratory function, which may have affected the quality of the EBC sample. The other patient was wearing a corset due to multiple bone metastases, which may have limited breathing during breath condensation collection. Based on the clinical records, this patient may have had undifferentiated large cell carcinoma rather than adenocarcinoma, and therefore the unusual conditions of these two patients may have influenced the results of the EBC analysis.

[0183] To evaluate the potential use of extracellular particles in EBCs for the diagnosis of non-squamous lung cancer, we developed and evaluated a diagnostic model using LOOCV and linear support vector classification (SVC). The model achieved a predictive accuracy of 0.89 (area under the receiver operating characteristic curve (AUC)) (Figure 12D). This suggests that extracellular particles in EBCs can be used to aid in the non-invasive diagnosis of non-squamous lung cancer. Next, we performed feature selection to identify particles that could function as biomarkers. SelectKBest analysis suggested that populations of CellMask-positive and Annexin V-positive particles were unique to non-squamous lung cancer patients (Figure 12E). These extracellular particle populations were elevated in EBCs from non-squamous lung cancer patients (Figure 12F), suggesting that they may be derived from adenocarcinoma cells. To confirm this, we examined the expression of EpCAM and podoplanin, markers of alveolar type II epithelial cells that are increased in cancer. We found that particles positive for these markers were enriched in non-squamous lung cancer samples (Figure 12G). A diagnostic model (two parameters) based on the concentrations of EpCAM-positive and podoplanin-positive extracellular microparticles achieved an AUC of 0.84, whereas the AUC of models based on three or two clinical markers was approximately 0.3 (Figures 12H and 12I).

[0184] 6. Conclusion In this study, we proposed a novel concept for staining all components of extracellular particles and demonstrated that comprehensive profiling and sorting of extracellular particles is possible using high-resolution flow cytometry. The simplified staining procedure without FSC-based thresholding and wash steps in the BD Influx system was preferable for this purpose. We developed a robust multicolor panel and implemented machine learning-based data processing to visualize the entire extracellular particle population, including non-membranous particles. Previous applications of flow cytometry have focused on specific extracellular particle populations, such as exosomes and viral particles. In contrast, this study enabled the simultaneous detection of not only extracellular vesicles and viral particles, but also chromatin, protozoa, and mitochondrial compartments. Therefore, this method provides more comprehensive insights into the field of biological particle research.

[0185] In this example, we analyzed extracellular microparticles in human BALF and EBCs to reveal specific extracellular microparticle populations and their interpatient diversity. In particular, this method is based on versatile staining targeting shared extracellular microparticle characteristics and does not require a washing procedure, making it easily applicable to many samples. This example suggests that extracellular microparticle phenotypes may be useful for predicting prognosis. Furthermore, specific extracellular microparticle populations were detected in EBCs, suggesting that they may be potential markers for non-squamous lung cancer. Non-squamous lung cancer is a subtype of non-small cell lung cancer (NSCLC) and is the most common form of lung cancer. Adenocarcinoma is the most common type of non-squamous NSCLC, accounting for approximately 40% of lung cancer cases. Unlike smoking-related small cell lung cancer (SCLC) and squamous NSCLC, adenocarcinoma also occurs in low-risk groups, such as non-smokers in women. However, there are currently no established methods for screening for non-squamous lung cancer in low-risk groups. Therefore, noninvasive screening for nonsquamous lung cancer from exhaled breath is a promising strategy. The extracellular particle population identified in this study was predicted to be derived from adenocarcinoma based on the expression of EpCAM and podoplanin, markers of alveolar type II epithelial cells that are increased in cancer (Figure 12G).

[0186] Furthermore, there is only one report on volatile organic compounds (VOCs)-based diagnosis of adenocarcinoma and squamous cell carcinoma, and the resolution of histological identification is insufficient compared to the accuracy of discrimination between healthy subjects and cancer patients. Therefore, there are currently no diagnostic techniques for identifying non-squamous cell lung cancer other than invasive biopsy. This example not only provides a breakthrough in the development of diagnostic techniques for non-squamous cell lung cancer, which has been difficult to identify non-invasively, but also suggests the potential for revealing factors related to pathophysiology that cannot be detected by VOC analysis or other methods. Furthermore, the method of this embodiment can detect not only host-derived particles but also pathogens such as virus particles and protozoa, making it possible to diagnose other diseases. [Explanation of symbols]

[0187] 1...information acquisition and analysis device, 2...detection unit, 3...information processing unit, 4...input unit, 5...output unit, 31...RAM, 32...ROM, 33...storage, 34...CPU, 35...receiving means, 36...transmitting means, 37...system bus.

Claims

1. 1. A method for aiding in the diagnosis or prognosis of a disease, comprising: From each subject in the population of subjects, (A) extracellular particle information on a population of extracellular particles collected from the lungs or exhaled breath of the subject, the extracellular particle information including, for each extracellular particle of the population of extracellular particles, scattered light information on scattered light of irradiated light and luminescence information on a plurality of luminescences with different peak wavelengths that are caused by the irradiated light from components of the extracellular particles or from labeling substances bound to the components; and (B) disease information regarding the disease status in the subject; and Obtaining information for assisting in the diagnosis or prognosis prediction of the disease associated with extracellular particles collected from the lungs or exhaled breath by machine learning based on a dataset including the extracellular particle information and the disease information; A method comprising:

2. the population of subjects comprises a first population of subjects affected with the disease and a second population of subjects not affected with the disease; The disease information includes information on whether the subject is suffering from the disease. The method of claim 1.

3. and acquiring characteristics of the extracellular microparticles characteristic of the first population through the machine learning. The method of claim 2.

4. the machine learning identifies markers for the disease; The method of claim 2.

5. creating a diagnostic model for the disease through the machine learning; The method of claim 2.

6. the population of subjects comprises a population of subjects suffering from a disease; the disease information includes a prognosis status of the subject; The method of claim 1.

7. A patient prognosis prediction model is created by the machine learning. The method of claim 6.

8. 8. A method according to claim 5 or 7, A model for diagnosing said disease or for predicting the prognosis of a patient.

9. the disease is a pulmonary disease; 9. The model of claim 8.

10. A diagnostic or prognostic model for non-squamous lung cancer, comprising: receiving extracellular particle information relating to a population of extracellular particles collected from the lungs or exhaled breath of a subject, the extracellular particle information including, for each extracellular particle of the population of extracellular particles, scattered light information relating to scattered light of irradiated light and luminescence information relating to a plurality of luminescences with different peak wavelengths that are caused by the irradiated light from components of the extracellular particles or labeled substances bound to the components; and diagnosing or predicting the prognosis of non-squamous lung cancer based on the concentration of at least one of EpCAM-positive extracellular microparticles and podoplanin-positive extracellular microparticles in the population of extracellular microparticles. Model.

11. 1. A method for analyzing a sample collected from the lungs or exhaled air and containing a population of extracellular particulates labeled with a plurality of labeling substances, comprising: acquiring extracellular particle information for each extracellular particle contained in the group of extracellular particles, the information including scattered light information regarding scattered light of the irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are generated from components of the extracellular particles or labeling substances bound to the components due to the irradiated light; analyzing the population of extracellular particles by machine learning based on the extracellular particle information; A method comprising:

12. analyzing the population of extracellular particulates generating a feature vector for each extracellular particle included in the group of extracellular particles based on the acquired extracellular particle information; mapping at least some of the plurality of extracellular particles contained in the group of extracellular particles in two or three dimensions based on the similarity of a representative vector consisting of some or all of the components of the feature vector, and labeling each of the mapped plurality of extracellular particles based on at least a plurality of components of the representative vector that correspond to the luminescence information, thereby creating a first scatter diagram; The method of claim 11 , comprising:

13. analyzing the population of extracellular particulates generating a feature vector for each extracellular particle included in the group of extracellular particles based on the acquired extracellular particle information; classifying the extracellular particles contained in the group of extracellular particles into two or more clusters based on the similarity of a representative vector consisting of some or all of the components of the feature vector; The method of claim 11 , comprising:

14. 1. A program for analyzing a sample collected from the lungs or exhaled breath of a subject, the sample including a population of extracellular particles labeled with a plurality of labeling substances, the program comprising: Computer, an extracellular particle information acquiring means for acquiring extracellular particle information for each extracellular particle contained in the group of extracellular particles, the extracellular particle information including scattered light information regarding scattered light of irradiated light and luminescence information regarding a plurality of luminescences with different peak wavelengths that are generated from components of the extracellular particles or labeling substances bound to the components due to the irradiated light; an analysis means for analyzing the population of extracellular particles by machine learning based on the extracellular particle information; A program that functions as a

15. Computers, and more a disease information acquiring means for acquiring disease information relating to the state of a disease in the subject; an auxiliary information generating means for generating information for assisting in the diagnosis or prognosis prediction of the disease associated with the extracellular particles collected from the lungs or exhaled breath by machine learning based on a dataset including the extracellular particle information and the disease information; The program according to claim 14,

16. Computers, and more The acquired extracellular particle information is input into a model created by the method according to claim 5 or 7 to produce a diagnostic model for diagnosing a disease or a prognosis prediction model for predicting the prognosis of the subject. The program according to claim 14,

Images