Method for preparing protein markers for detecting atopic dermatitis

Abstract

To provide a method for detecting atopic dermatitis in a low-invasive or non-invasive manner.SOLUTION: The present invention collects protein markers for detecting atopic dermatitis from skin surface lipids collected from subjects. The present invention detects atopic dermatitis using the protein markers.SELECTED DRAWING: None

Description

[Technical Field]

[0001] The present invention relates to a method for preparing a protein marker for detecting atopic dermatitis, and The present invention relates to a method for detecting atopic dermatitis using a quality marker. [Background technology]

[0002] Atopic dermatitis (hereinafter also referred to as "AD") ) is an eczematous skin disease that occurs mainly in people with a predisposition to atopy. Symptoms include chronic and recurrent itching, rash, erythema, etc., occurring bilaterally and contralaterally, as well as hypokeratosis. , impaired barrier function, dry skin, etc. AD mostly occurs in infants and young children, and tends to improve as they grow older. However, in recent years, adult-onset and refractory AD have also been increasing.

[0003] Newborns or infants who have a genetic predisposition to allergies or atopy should Rashes, atopic dermatitis, food allergies, as well as bronchial asthma and allergic rhinitis, It has been reported that various allergic diseases develop with age (allergic march). In this way, when one allergic disease develops, it is possible to develop another allergic disease. Therefore, it is important to take care of your child in the early stages of their life. There is a need to prevent the onset of allergic diseases.

[0004] Various evaluation methods have been used in clinical settings to assess the pathology of AD. Dermatologists evaluated the severity of atopic dermatitis using the Severity Scoring of Atopy. c Dermatitis(SCORAD) and Eczema Area and Sev However, these evaluation methods are Because it is highly subjective, it is used in conjunction with more objective biomarkers. Serological evaluation items for AD include total IgE levels in the blood, peripheral blood eosinophil counts, Serum lactate dehydrogenase (LDH) level, serum thymus a nd activation-regulated chemokine (TARC) value However, serological evaluation is invasive because it requires blood sampling. Furthermore, it cannot necessarily be said that diagnosis is possible with sufficient accuracy.

[0005] Skin is a tissue that can be used to collect biological samples minimally invasively because it is in contact with the outside world. Various tests have been carried out using skin samples such as biopsy skin tissue and tape-stripped stratum corneum. Nucleic acid or protein markers have been isolated. Non-Patent Documents 1 to 6 and Patent Document 1 include: By applying a weak adhesive tape to the skin, interleukin 1 is non-invasively delivered from the skin surface. Interleukins (ILs), TNF-α, INF-γ, human β-defensin (hBD2), and other peptides It is described that peptide markers were collected and used to examine skin diseases and conditions. Patent Document 2 describes a method for extracting RNA and the like derived from skin cells of a subject from lipids on the skin surface. It is described that nucleic acids are isolated and used as samples for biological analysis. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] International Publication No. 2014 / 144289 [Patent Document 2] International Publication No. 2018 / 008319 [Non-patent literature]

[0007] [Non-Patent Document 1] Skin Res Technol, 2001, 7(4):227-37 [Non-patent document 2] Skin Res Technol, 2002, 8(3):187-93 [Non-patent document 3] Med Devices (Auckl), 2016, 9:409-417 [Non-patent document 4] Med Devices (Auckl), 2018, 11:87-94 [Non-patent document 5] J Tissue Viability, 2019, 28(1):1-6 [Non-patent document 6] J Diabetes Res, doi / 10.1155 / 2019 / 1973704 Summary of the Invention [Problem to be solved by the invention]

[0008] The present invention provides a method for minimally or non-invasively extracting a protein marker for detecting atopic dermatitis from a subject. and a method for detecting atopic dermatitis using said protein marker. Regarding. [Means for solving the problem]

[0009] In one aspect, the present invention provides a method for determining lipids on the skin surface from lipids collected from a subject, using the following Tables 1-1 to 1-1. -13. At least one protein selected from the group consisting of the proteins shown in The present invention provides a method for preparing a protein marker for detecting atopic dermatitis, comprising: In another aspect, the present invention provides a method for determining lipids on the skin surface collected from a subject, the lipids being determined from the lipids in Table 1-1 below. At least one protein selected from the group consisting of the proteins shown in 1-13 is detected. The present invention provides a method for detecting atopic dermatitis in a subject, the method comprising: In yet another aspect, the present invention provides a method for producing a pharmaceutical composition comprising the proteins shown in Tables 2-1 to 2-5 below. A protein for detecting atopic dermatitis, comprising at least one protein selected from the group consisting of: Provides quality markers. [Effects of the Invention]

[0010] According to the present invention, atopic skin can be extracted from a subject by a simple and minimally invasive or non-invasive method. A protein marker for detecting atopic dermatitis is collected, or the marker is used to detect atopic dermatitis. Therefore, the present invention can be applied to a case where invasive biological sample collection is not easy. The present invention also enables the diagnosis of atopic dermatitis in a variety of subjects, including infants. The method may contribute to the early diagnosis and treatment of atopic dermatitis in infants and adults. DETAILED DESCRIPTION OF THE INVENTION

[0011] All patents, non-patent documents, and other publications cited herein are hereby incorporated by reference in their entirety. The bodies are incorporated herein by reference.

[0012] The names of the proteins disclosed herein are derived from UniProt (https: / / www.uniprot Follow the Gene Name or Protein Name listed in the [http: / / www.ncbi.nlm.nih.gov / ].

[0013] In this specification, the term "infant" refers to a child of age from 0 years old to the age of entering school, specifically, a child of age 0 years old. In this specification, "adult" refers to a person between the ages of 12 and 5. In a broad sense, "adult" also refers to "infants." It refers to people who have not undergone puberty, but preferably refers to people who have completed secondary sexual characteristics, specifically those aged 16 or over. Preferably, people aged 20 or over are preferred.

[0014] In this specification, "skin surface lipids (SS) "L" refers to the fat-soluble fraction present on the surface of the skin, sometimes called sebum. In addition, SSL mainly contains secretions from exocrine glands such as sebaceous glands in the skin, and It is present on the surface of the skin in the form of a thin layer that covers the surface.

[0015] In this specification, unless otherwise specified, the term "skin" refers to the epidermis, dermis, hair follicles, etc. of the body surface. It is a general term for the area that includes tissues such as sweat glands, sebaceous glands, and other glands.

[0016] In the present invention, "detection" of atopic dermatitis means the detection of atopic dermatitis in a subject from whom an SSL was collected. It is meant to clarify the presence or absence of atopic dermatitis, and is used for testing, measuring, judging, and evaluating Alternatively, it can be expressed as the term "evaluation support." The term "assessment" does not include medical assessment or evaluation.

[0017] In this specification, the term "feature" is synonymous with the term "explanatory variable" in machine learning. In the specification, the protein markers used for machine learning selected from the protein markers for detecting atopic dermatitis are The proteins used are sometimes called "feature proteins."

[0018] The present inventors have discovered that SSL contains proteins useful for detecting atopic dermatitis (AD). These proteins are useful as protein markers for detecting AD. Therefore, the SSL can be conveniently collected from the skin surface of the subject. and the detection of AD in a subject by minimally invasive or non-invasive methods. The protein markers involved can be recovered.

[0019] Therefore, in one aspect, the present invention provides protein markers for detecting AD. In one embodiment, the present invention provides a method for preparing a protein marker for detecting AD. The method comprises extracting a target protein marker for detecting AD from SSL collected from a subject. In another embodiment, the present invention provides a method for detecting AD. The method includes detecting the protein marker for detecting AD from SSL collected from a subject. Includes.

[0020] The method for preparing a protein marker for detecting AD and the method for detecting AD according to the present invention (hereinafter referred to as "Summary"). In the method of the present invention, the subject is an organism having SSL on its skin. Examples of subjects include mammals, including humans and non-human mammals, and preferably The subject is preferably a human. The subject may be of any gender or age, and may range from infants to adults. Preferably, the subject is a mammal in need of or desiring detection of AD (preferably For example, the subject is a mammal (preferably a human) suspected of developing AD. )

[0021] In one embodiment, the method further comprises harvesting the subject's SSL. The skin from which SSL is collected may be from the head, face, neck, trunk, limbs, etc. The skin of any part of the body can be used, preferably a part of the body with symptoms such as AD-like eczema or dryness. Skin is one example.

[0022] The collection of SSLs from the subject's skin involves the use of a device to retrieve or remove the SSLs from the skin. Any means that can be used for the absorption of the material may be employed. Preferably, the material is an SSL absorbent material, S SL adhesive material or a tool to scrape SSL from the skin can be used. As for absorbent materials or SSL adhesive materials, materials that have affinity for SSL are particularly suitable. Examples of the material include, but are not limited to, polypropylene and pulp. A more detailed example of the order is the SS on sheet-like materials such as oil blotting paper and oil blotting film. How to absorb L, how to attach SSL to glass plates, tape, etc., how to use a spatula, a scraper, etc. SSL is scraped off and collected by a rapist. To improve adhesion, SSL absorbent materials pre-soaked with highly lipid-soluble solvents may be used. On the other hand, SSL absorbent materials may not adsorb well if they contain highly water-soluble solvents or moisture. Therefore, it is preferable to use solvents with high water solubility and low water content. The material is preferably used in a dry state.

[0023] The collected SSL may be used immediately in the protein extraction step described below, but the protein may be The SSL may be stored until it is used in the protein extraction process. If stored, the SSL should be stored as soon as possible after collection. It is recommended to store SSL at a low temperature as soon as possible. Preferably, the temperature is between -20±20°C and -80±20°C, more preferably between -20±10 ° C. to -80±10° C., more preferably -20±20° C. to -40±20° C., and even more preferably Preferably, the temperature is -20±10°C to -40±10°C, more preferably -20±10°C, and even more preferably The storage period of the SSL is preferably within 12 months, but is not particularly limited thereto. months or less, for example, 6 hours or more and 12 months or less, more preferably 6 months or less, for example, 1 day or less It is preferably from 3 days to 6 months, more preferably from 3 months to 3 months, for example.

[0024] Protein extraction from collected SSL requires the use of techniques similar to those used for protein extraction or purification from biological samples. The method commonly used for preparation, e.g., water, phosphate-buffered saline ne solution, or a solution containing Triton X-100, Tween 20, etc. as a surfactant Extraction with liquid or M-PER buffer (Thermo Fisher Scientific) Scientific), MPEX PTS Reagent(GL science) , QIAzol Lysis Reagent (Qiagen), EasyPep TM M ini MS Sample Prep Kit(ThermoFisher Scie Commercially available protein extraction reagents and kits such as ntific can be used. Cut.

[0025] The extracted SSL-derived proteins contain one protein marker for AD detection, which will be described later. The SSL-derived protein may be used immediately for the detection of AD, as described below. However, it may be stored under normal protein storage conditions until it is used for detecting AD.

[0026] As shown in the Examples below, 418 SSL-derived proteins shown in Tables 1-1 to 1-13 Proteins whose abundance in SSLs is significantly altered in AD patients compared with healthy controls In addition, we constructed a model using machine learning that uses the abundance of these proteins in the SSL as a feature. The constructed prediction model makes it possible to predict AD. The SSL-derived proteins shown can be used as protein markers for AD detection. Among the proteins shown in Tables 1-1 to 1-13, 147 types shown in Tables 2-1 to 2-5 were identified. As will be shown in the examples below, proteins are new proteins that have not been reported to be related to AD. It is a novel protein marker for detecting AD. The proteins are 200 types shown in Tables 4-1 to 4-6 below and 5-1 to 5-6 below. 9, which is a protein common to the 283 proteins shown in It can be preferably used as a car.

[0027] [Table 1-1]

[0028] [Table 1-2]

[0029] [Table 1-3]

[0030] [Table 1-4]

[0031] [Table 1-5]

[0032] [Table 1-6]

[0033] Table 1-7

[0034] Table 1-8

[0035] Table 1-9

[0036] Table 1-10

[0037] Table 1-11

[0038] Table 1-12

[0039] Table 1-13

[0040] Table 2-1

[0041] Table 2-2

[0042] Table 2-3

[0043] [Table 2-4]

[0044] [Table 2-5]

[0045] [Table 3-1]

[0046] [Table 3-2]

[0047] More specifically, the SSL-derived proteins shown in Tables 1-1 to 1-13 are the same as those in Tables 4-1 to 4-4 below. -6 and the proteins shown in Tables 5-1 to 5-9, and the proteins shown in Tables 4-1 to 4-6 Refer to the following Tables 7-1 to 7-4, Table 8, Tables 11-1 to 11-4, Tables 12-1 to 12-4 and Table 13, and the proteins shown in Tables 5-1 to 5-9 are listed in Tables 9-1 to 9 below. -7, Tables 10-1 to 10-2, Tables 14-1 to 14-7, Tables 15-1 to 15-4 and Table 16 The protein includes the protein shown in

[0048] As shown in the examples below, the results were extracted from the SSLs of healthy children and children with AD. is the quantitative value of proteins that were quantified in 75% or more of subjects in either group of AD children. The results showed that the abundance ratio was 1.5 times higher in children with AD than in healthy children (p≦0. 05), and 116 proteins (Tables 7-1 to 7-4) increased by 0.75 times or less (p 12 proteins (Table 8) that were decreased to a level of 0.05 or less were identified. and adult AD patient 2 SSL extracted from the group of healthy subjects or AD patients, and more than 75% The quantitative values ​​of proteins obtained in the subjects were analyzed. The abundance ratio of 205 proteins ( Tables 9-1 to 9-7), and 37 proteins that were reduced by 0.75-fold or less (p≦0.05) The qualities (Tables 10-1 to 10-2) were identified.

[0049] Therefore, in one embodiment, the method for detecting AD of the present invention comprises detecting a leukemia in the SSL of a subject. AD is detected based on the amount of the protein marker for detecting AD (for example, the concentration of the marker in SSL). This includes detecting.

[0050] For example, Tables 7-1 to 7-4, Table 8, Tables 9-1 to 9-7, and Table 1 in the subject SSL The SSL is derived based on the concentration of one or more protein markers shown in 0-1 to 10-2. Whether the subject has AD (in other words, whether the SSL is derived from a subject with AD) In the AD detection method of the present invention, the AD detection method shown in Tables 7-1 to 7-4 , any one of the proteins shown in Table 8, Tables 9-1 to 9-7, and Tables 10-1 to 10-2 or a combination of two or more of these proteins as a protein marker for detecting AD. For example, the one or more markers (target markers) in the SSL of the subject can be detected. ) and comparing the measured marker concentrations with those of a healthy group, It can detect whether or not a person has AD.

[0051] The target marker is selected from the group consisting of proteins shown in Tables 7-1 to 7-4 and Tables 9-1 to 9-7. When the target marker is at least one selected from the above, the concentration of the target marker in the subject is If the level of the target marker in the subject is higher than that of the target marker in the subject, the subject can be detected as having AD. If the concentration of CAR is statistically significantly higher than that of a healthy group, the subject can be detected as having AD. For example, the concentration of the target marker in a subject is preferably 11 times higher than that in a healthy group. If the ratio is 0% or more, more preferably 120% or more, and even more preferably 150% or more, the The subject can be detected as having AD. Two or more protein markers for detecting AD can be used as target markers. When using a marker, a certain percentage of the target marker, for example, 50% or more, preferably 70% or more, is used. Whether or not the above criteria are met, more preferably 90% or more, and even more preferably 100% Based on this, AD can be detected in a subject.

[0052] The target marker is selected from the group consisting of proteins shown in Table 8 and Tables 10-1 to 10-2. When the target marker is at least one of the above, the concentration of the target marker in the subject is higher than that of a healthy group. If the concentration of the target marker in the subject is low, the subject can be detected as having AD. If the level is statistically significantly lower than that of a healthy group, the subject can be detected as having AD. For example, the concentration of the target marker in the subject is preferably 90% or less of that in a healthy group, More preferably, the subject is diagnosed with AD if the percentage is 80% or less, and even more preferably, 75% or less. When two or more protein markers for detecting AD are used as target markers, In this case, a certain percentage of the target markers, for example, 50% or more, preferably 70% or more, more preferably The affected individuals are selected based on whether or not at least 90%, and more preferably 100%, meet the above criteria. AD can be detected in a subject.

[0053] The concentration of protein markers for detecting AD in the SSL was measured by ELISA, immunohistochemistry, fluorescence, and electrophoresis. Using conventional protein detection or quantification methods such as electrophoresis, chromatography, and mass spectrometry. Among these, mass spectrometry such as LC-MS / MS is preferred. For concentration measurement, the above SSL-derived protein was used as a sample, and the target was measured according to the usual procedure. The detection or quantification of one or more protein markers that are the target of the calculation may be carried out. The concentration of a target marker may be based on the absolute amount of the target marker in the SSL or on the amount of the target marker in the SS. It may be the relative concentration to other standards or to the total protein in L.

[0054] The healthy group may be a population not affected by AD. The healthy group may be a population of the same species as the subject. If necessary, a group that constitutes a healthy group should be selected according to the characteristics of the subjects. For example, if the subjects are infants, a group of healthy infants may be used as the healthy group. Or, if the subjects are adults, a healthy adult population may be used as the healthy group. The concentration of the protein marker for detecting AD in the healthy group can be determined by the above-mentioned procedure. Preferably, the concentration of the marker in a healthy group has been measured in advance. More preferably, the results of the tests in Tables 7-1 to 7-4, Table 8, Tables 9-1 to 9-7, and The concentrations of all the markers shown in Tables 10-1 to 10-2 have been measured in advance.

[0055] Alternatively, a protein selected from the group consisting of proteins shown in Tables 7-1 to 7-4 and Tables 9-1 to 9-7. and a group consisting of at least one selected from the group consisting of proteins shown in Table 8 and Tables 10-1 to 10-2. The target marker may be a combination of at least one selected from the following: AD The criteria for detection are the same as above.

[0056] In one embodiment of the method for detecting AD of the present invention, when the subject is an infant, the target marker is A small number of protein markers selected from the protein markers for detecting AD shown in Tables 7-1 to 7-4 and Table 8 When the subject is an adult, the target markers are those listed in Tables 9-1 to 9-7 and Table 1 At least one protein marker for detecting AD selected from the protein markers shown in 0-1 to 10-2 Seeds are preferred.

[0057] Other preferred examples of protein markers for detecting AD in infants include those listed in Table 11-1 below. The proteins shown in Tables 11-1 to 11-4 are 127 types. Proteins were extracted from the SSL of healthy and AD-affected children, and quantitation was observed in more than 75% of all subjects. Among the proteins for which the expression was obtained, the abundance ratio in AD children was 1.5 times or more (p ≦0.05) or decreased 0.75-fold or less (p≦0.05). Other preferred examples of protein markers for detecting AD in adults include those listed in Tables 14-1 to 14-2 below. The 220 proteins shown in Tables 14-1 to 14-7 are preferred examples. The proteins shown in 7 were extracted from the SSLs of healthy adults and adult AD patients, and the Among the proteins for which quantification values ​​were obtained for 75% or more, the abundance ratio was higher in AD patients compared to healthy individuals. Increased by 1.5 times or more (p≦0.05) or decreased by 0.75 times or less (p≦0.05) It is protein.

[0058] Therefore, in another embodiment of the method for detecting AD of the present invention, when the subject is an infant, The target markers are selected from the protein markers for detecting AD shown in Tables 11-1 to 11-4. When the subject is an adult, the target marker is preferably one selected from Table 14-1 to At least one protein marker for detecting AD is preferably selected from the protein markers for detecting AD shown in 14-7. Alternatively, if the subjects include both infants and adults, see Tables 11-1 to 11-4. and at least one protein selected from the group consisting of proteins shown in Tables 14-1 to 14-7. and at least one protein selected from the group consisting of the target marker It may also be used as.

[0059] In a further embodiment, the method for detecting AD of the present invention comprises detecting said AD in the SSL of a subject. Prediction was constructed using the amount of protein marker to be detected (e.g., marker concentration in SSL). For example, the protein marker for detecting AD is The amount of AD is used as the explanatory variable, and whether or not it is AD is used as the objective variable, and the machine learning algorithm is used to analyze AD. A discriminant (prediction model) is constructed to distinguish between patients and healthy individuals. The amount (concentration) of the marker can be either an absolute value or a relative value. Alternatively, the data may be normalized.

[0060] In one embodiment, the quantitative values ​​of the target markers derived from the SSL of an AD patient and the SSL of a healthy subject are compared. Using the quantitative values ​​of the target markers derived from the markers as training samples, a discriminant ( A predictive model is constructed and used to distinguish between atopic dermatitis patients and healthy individuals based on the discriminant equation. The cutoff value (reference value) is then determined. The target marker is then detected from the SSL collected from the subject. The amount of the test substance is measured, the measured value is substituted into the discriminant, and the result obtained from the discriminant is used as a reference. By comparing with a control value, the presence or absence of AD in a subject can be detected.

[0061] The variables used to construct the discriminant equation are explanatory variables and target variables. The expression level of the protein marker for detecting AD selected by the following method (e.g., SSL concentration) The objective variable can be, for example, whether the sample is from a healthy individual or D Patient-derived (with or without AD) can be used.

[0062] The features include statistically significant differences between the two groups to be discriminated, e.g., differences in expression levels between the two groups. Proteins that fluctuate significantly (expression-varying proteins) are used as feature proteins, and their expression levels are In addition, the feature proteins can be generated by known methods such as machine learning algorithms. For example, the following random forest algorithm can be used to extract the We used the expression levels of proteins with high variable importance in the dataset, and It is possible to extract protein features using the "a" package.

[0063] As the algorithm for constructing the discriminant, publicly available algorithms such as those used for machine learning are used. Examples of machine learning algorithms include Random Forest. Random forest, linear kernel support vector machine (SVM) linear), rbf kernel support vector machine (SVM rbf) Neural net, Generalized Linear Model d linear model), regularized linear discriminant analysis (Regularized li near discriminant analysis), regularized logistic regression (Regularized logistic regression) The validation data is input into the constructed prediction model to calculate a predicted value, and the predicted value is compared with the actual measurement value. The model that best fits the model, e.g., the model with the highest accuracy, is selected as the optimal model. It can be selected as a prediction model. In addition, the detection rate (Recalculated) can be calculated from the predicted value and the actual measured value. l), precision, and their harmonic mean F-measure, and The model with the highest value can be selected as the optimal predictive model.

[0064] When using the random forest algorithm to build the discriminant, the predictive model As an index of the accuracy of the estimation, the OOB error rate ) can be calculated (Breiman L. Machine Learning (2001) 45;5-32). In the forest, we use a method called bootstrap to extract overlaps from all samples. By allowing the sample to be randomly extracted, about two-thirds of the total number of samples, a classification called a decision tree is created. At this time, samples that are not extracted are marked as Out of Bug (OOB). A single decision tree is used to predict the OOB target variable and compare it with the correct label. By doing so, the error rate can be calculated (OOB error rate in decision tree). The same process was repeated 500 times to calculate the OOB error for 500 decision trees. The average value of the OOB error rate is used as the OOB error rate for the random forest model. or rate.

[0065] The number of decision trees (ntree value) used to build the random forest model is set to the default. The default is 500, but this can be changed to any number as needed. The number of variables (metric values) used to create the discriminant equation for a sample in one decision tree is By default, this is the square root of the number of explanatory variables, but you can change it to one or all explanatory variables as needed. The mtry value can be changed to any value up to the number of You can use the "ret" package. Specify the dam forest and try eight different mtry values. For example, The mtry value can be selected as the optimum mtry value. The number of trials can be changed to any number of trials as needed.

[0066] When using the random forest algorithm to build the discriminant, the model structure The importance of the explanatory variables used in the construction can be expressed as a numerical value (variable importance). For example, the decrease in the Gini coefficient (Mean Decrease Gini) can be used to This can be done.

[0067] The method for determining the cutoff value (reference value) is not particularly limited, and can be determined according to a known method. For example, the ROC (Receiver Operator Criteria) created using the discriminant It can be obtained from the (Lattice Characteristic Curve) In an ROC curve, the vertical axis represents the probability of a positive result in a positive patient (sensitivity) and the horizontal axis represents the probability of a negative result. The false positive rate is calculated by subtracting the probability of a negative result (specificity) from 1 in a patient. The ROC curve shows "true positive (sensitivity)" and "false positive (1-specificity)". Regarding the sensitivity, the value at which "true positive (sensitivity)" - "false positive (1 - specificity)" is maximized (Youden The index) can be used as a cutoff value (reference value).

[0068] When using data on a large number of proteins to build a predictive model, it is necessary to You can also compress the data using PCA before building a predictive model. For example, we perform dimensionality reduction by principal component analysis of the quantitative values ​​of proteins, and then use the main principal components to create a prediction model. These can be used as explanatory variables for constructing the model.

[0069] As shown in the Examples below, changes in expression were observed between healthy children and children with AD. Proteins 1-4 were used as feature proteins, and their quantitative data (Log2(Abunda The explanatory variables were the nce+1) value, and the target variables were the healthy children and the AD-affected children. Random forest (Breiman L. Machine Learning (2001) 45;5-32) was used as the rhythm. The constructed prediction model makes it possible to predict AD in infants and young children. Furthermore, as shown in the Examples below, it was shown that the α-glucan is expressed in healthy adults and adult AD patients. For the proteins in Tables 14-1 to 14-7 where fluctuations were observed, a similarly constructed prediction model was used. It has been shown that the prediction of adult AD is possible by the model. In one embodiment, the subject is an infant and the target marker is selected from the group consisting of those listed in Tables 11-1 to 11-4. In another embodiment of the method for detecting AD of the present invention, The subjects were adults, and the target markers were 220 proteins shown in Tables 14-1 to 14-7. It's quality.

[0070] As shown in the examples below, healthy children and children with AD were used as subjects, and SSL from the subjects was obtained. The quantitative data of the derived protein (Log2(Abundance+1) value) was used as an explanatory variable. The objective variables were healthy children and children with AD, and random forest was used as a machine learning algorithm. We attempted to extract feature proteins and build a prediction model using the algorithm. The top 140 proteins with variable importance based on the Gini coefficient (Tables 12-1 to 12-2) -4) was selected as a feature protein and used to build a prediction model. It was shown that the prediction model can predict AD in infants. As shown, healthy adults and AD patients were used as subjects, and SSL-derived samples from the subjects were analyzed. Similarly, using the protein quantitative data (Log2(Abundance+1) value), We attempted to extract trace proteins and build a prediction model. The top 110 proteins (Tables 15-1 to 15-4) were selected as feature proteins. This was used to construct a prediction model. The constructed prediction model makes it possible to predict adult AD. Therefore, in one embodiment of the method for detecting AD of the present invention, the subject The target markers are 140 proteins shown in Tables 12-1 to 12-4. In another embodiment of the method for detecting AD of the present invention, the subject is an adult, and the target marker The proteins are 110 types shown in Tables 15-1 to 15-4.

[0071] As will be shown in the examples below, healthy children and children with AD were used as subjects, and SSL-derived The quantitative data of the protein (Log2(Abundance+1) value) was used as an explanatory variable. The objective variables were healthy children and children with AD, and the Boruta method (Kurs a et al. Fundamental Informaticae (2010) 101;271-286) to identify the protein features. Extraction was performed (maximum number of trials: 1000, p-value less than 0.01). 35 proteins (Table 13) were extracted as feature proteins. The quantitative data of these proteins were used as feature proteins. A predictive model constructed using random forests can predict AD in infants and young children. As will be shown in the examples below, it was shown that healthy adults and AD patients The subjects were used as subjects, and the quantitative data (Log2(Abunda)) of SSL-derived proteins from the subjects were used. nce+1) value) was used as an explanatory variable to similarly extract feature proteins. The proteins (Table 16) were extracted as feature proteins. A predictive model constructed using DamForest was shown to be capable of predicting adult AD. Therefore, in another embodiment of the method for detecting AD of the present invention, the subject is an infant. The protein markers for detecting AD are the 35 proteins shown in Table 13. In another embodiment of the present method for detecting AD, the subject is an adult, and the AD detection protein The protein markers are the 24 proteins shown in Table 16.

[0072] Among the above-mentioned protein markers for detecting AD, Tables 7-1 to 7-7 were extracted by expression analysis. 130 proteins (A) included in any of Tables 11-4, 11-8, and 11-1 to 11-4; Tables 12-1 to 12-2 selected as feature proteins by random forest 4 and the 140 proteins (B) shown in Fig. 4 were used as feature proteins by the Boruta method. The union (A∪B∪C) of the 35 proteins (C) selected in Table 13 is shown in Table 4-1 The proteins are listed in Tables 4-1 to 4-6. At least one protein selected from the group consisting of the following is a preferred example of an anti-infantile AD protein according to the present invention: The amount of the at least one protein is used as a detection marker. Infant AD can be detected by comparing the at least A prediction model is constructed using one type of protein as a feature protein, and This can detect AD in infants.

[0073] [Table 4-1]

[0074] [Table 4-2]

[0075] [Table 4-3]

[0076] [Table 4-4]

[0077] [Table 4-5]

[0078] [Table 4-6]

[0079] Among the proteins shown in Tables 4-1 to 4-6 above, POF1B (Protein POF1B), MNDA (My eloid cell nuclear differentiation antigen), SERPINB4 (Serpin B4), CLEC3B (Tet ranectin), PLEC (Plectin), LGALS7 (Galectin-7), H2AC4 (Histone H2A type 1-B / E ), SERPINB3 (Serpin B3), AMBP (Protein AMBP), PFN1 (Profilin-1), DSC3 (Desmo collin-3), IGHG1(Immunoglobulin heavy constant gamma 1), ORM1(Alpha-1-acid g lycoprotein 1), RECQL (ATP-dependent DNA helicase Q1), RPL26 (60S ribosomal pr otein L26), KLK13 (Kallikrein-13), RPL22 (60S ribosomal protein L22), APOA2 ( Apolipoprotein A-II), SERPINB5 (Serpin B5), LCN15 (Lipocalin-15), IGHG3 (Immu noglobulin heavy constant gamma 3), CAP1 (Adenylyl cyclase-associated protein 1) ) and SPRR2F (Small proline-rich protein 2F) are the 23 proteins These 23 proteins are common to proteins (A), (B), and (C). In the present invention, at least one protein selected from the group consisting of The at least one protein is preferably used as a marker for detecting AD in infants and young children. By comparing the amount of ATP between subjects and healthy controls, infant AD can be detected. Alternatively, a prediction model is constructed using the at least one protein as a feature protein, Based on the prediction model, infantile AD can be detected.

[0080] In a preferred embodiment of the present invention, the two SSLs obtained from an infant subject are At least one, preferably two or more, selected from the group consisting of three types of proteins Preferably 5 or more proteins, more preferably 10 or more proteins, and even more preferably all proteins Furthermore, in the present invention, a small number of proteins selected from the group consisting of the 23 proteins are quantified. In addition to at least one protein, 200 proteins shown in Tables 4-1 to 4-6 below are also included. One or more proteins selected from the group consisting of proteins (excluding the above 23 proteins) For example, at least one selected from the group consisting of the 23 proteins may be quantified. In addition to one protein, 127 proteins shown in Tables 11-1 to 11-4 (previous At least one protein selected from the group consisting of the proteins listed in Table 12-1 to A group consisting of 140 proteins (excluding the above 23 proteins) shown in 12-4 and / or at least one of the 35 proteins shown in Table 13 (the above At least one protein selected from the group consisting of 23 proteins) may be quantified. When selecting proteins from Tables 11-1 to 11-4, the significance of the expression change should be considered. It is also possible to select preferentially from those with higher p-values ​​(e.g., smaller p-values). When selecting proteins from -1 to 12-4, proteins with higher variable importance are selected first. Proteins that are preferentially selected or ranked in the top 50, preferably 30, by variable importance The amount of at least one protein as described above may be selected from the group consisting of a subject and a healthy control group. By comparing the above, infant AD can be detected. A prediction model is constructed using at least one protein as a feature protein, and the prediction model AD can be detected in infants based on the results.

[0081] Among the above-mentioned protein markers for detecting AD, Tables 9-1 to 9 were extracted by expression analysis. -7, 242 proteins shown in Tables 10-1 to 10-2 and Tables 14-1 to 14-7 (D ) and Tables 15-1 to 15- 110 proteins (E) shown in 4 and the Boruta method were used to identify the protein features. The union (D∪E∪F) of the 24 proteins (F) selected in Table 16 is shown in Table 5-1 The proteins are listed in Tables 5-1 to 5-9. At least one protein selected from the group consisting of the following is a preferred example of a protein for use in the adult AD test of the present invention: The amount of the at least one protein is used as a protein marker for detection. By comparing the results of the analysis between the subjects and healthy controls, adult AD can be detected. A prediction model is constructed using at least one protein as a feature protein, and the prediction model Based on this, adult AD can be detected.

[0082] [Table 5-1]

[0083] [Table 5-2]

[0084] [Table 5-3]

[0085] [Table 5-4]

[0086] [Table 5-5]

[0087] [Table 5-6]

[0088] [Table 5-7]

[0089] [Table 5-8]

[0090] [Table 5-9]

[0091] Among the proteins shown in Tables 5-1 to 5-9 above, SERPINB1 (Leukocyte elastase inhibitor) ibitor), TTR (Transthyretin), DHX36 (ATP-dependent DNA / RNA helicase DHX36), I TIH4 (Inter-alpha-trypsin inhibitor heavy chain H4), GC (Vitamin D-binding prot ein), ALB (Serum albumin), SERPING1 (Plasma protease C1 inhibitor), DDX55 (AT P-dependent RNA helicase DDX55), IGHV1-46 (Immunoglobulin heavy variable 1-46) , EZR (Ezrin), VTN (Vitronectin), AHSG (Alpha-2-HS-glycoprotein), HPX (Hemope xin), PPIA (Peptidyl-prolyl cis-trans isomerase A), KNG1 (Kininogen-1), FN1 ( Fibronectin), PLG (Plasminogen), PRDX6 (Peroxiredoxin-6) and FLG2 (Filaggrin-2) ) are common to the above proteins (D), (E) and (F). At least one protein selected from these 19 proteins In the present invention, the protein is more preferably used as a marker for detecting adult AD. Detect adult AD by comparing the levels of at least one protein between subjects and healthy controls. Alternatively, the at least one protein can be used as a feature protein. A predictive model can be constructed and adult AD can be detected based on the predictive model.

[0092] In a preferred embodiment of the present invention, the 19 At least one, preferably two or more, more preferably Preferably, 5 or more kinds, more preferably 10 or more kinds, and even more preferably all kinds of proteins are included. In the present invention, at least one protein selected from the group consisting of the 19 proteins is quantified. In addition to one protein, 283 proteins shown in Tables 5-1 to 5-9 below were also included. At least one protein selected from the group consisting of (excluding the above 19 proteins) For example, the quality of at least one protein selected from the group consisting of the 19 proteins may be quantified. In addition to one protein, 220 proteins shown in Tables 14-1 to 14-7 at least one protein selected from the group consisting of: 110 proteins shown in 1 to 15-4 (excluding the above 19 proteins) and / or at least one selected from the group consisting of the 24 proteins shown in Table 16 ( Even if at least one protein selected from the group consisting of the above 19 proteins is quantified, In this case, when selecting proteins from Tables 14-1 to 14-7, it is recommended to consider the significance of expression changes. It is also possible to select with priority from those with higher significance (for example, smaller p-values). When selecting proteins 15-1 to 15-4, proteins with higher variable importance are selected. The top 50, preferably 30, variables are selected preferentially from the top. The amount of at least one protein may be selected from a group of proteins. By comparing the at least one type of test, adult AD can be detected. A prediction model was constructed using the protein as a feature protein, and adult A was predicted based on the prediction model. D can be detected.

[0093] As exemplary embodiments of the present invention, the following materials, manufacturing methods, uses, methods, etc. are further described herein. However, the present invention is not limited to these embodiments.

[0094] [1] Proteins shown in Tables 1-1 to 1-13 above were extracted from lipids on the skin surface collected from the subject. A method for treating atopic dermatitis, comprising recovering at least one protein selected from the group consisting of: Method for preparing a protein marker for detecting allergic dermatitis. [2] Proteins shown in Tables 1-1 to 1-13 above were extracted from lipids on the skin surface collected from the subject. The method comprises detecting at least one protein selected from the group consisting of Method for detecting atopic dermatitis in humans. [3] Preferably, the at least one protein is: At least one protein selected from the group consisting of proteins shown in Tables 2-1 to 2-5 Is it high quality? At least one protein selected from the group consisting of proteins shown in Tables 3-1 to 3-2 It is of high quality, The method described in [1] or [2]. [4] The subject is preferably an infant, the at least one protein Preferably, at least one protein selected from the group consisting of proteins shown in Tables 4-1 to 4-6. is also a type of protein, More preferably, the protein is selected from the group consisting of proteins shown in Tables 7-1 to 7-4 and Table 8. at least one protein More preferably, the protein is selected from the group consisting of proteins shown in Tables 11-1 to 11-4. or at least one protein listed in Tables 12-1 to 12-4. or at least one protein selected from the group consisting of the proteins shown in Table 13. at least one protein selected from the group consisting of proteins, More preferably, POF1B, MNDA, SERPINB4, CLEC3B, PLEC, LGALS7, H2AC4, SERPIN B3, AMBP, PFN1, DSC3, IGHG1, ORM1, RECQL, RPL26, KLK13, RPL22, APOA2, SERPINB5, At least one protein selected from the group consisting of LCN15, IGHG3, CAP1 and SPRR2F Including, More preferably, POF1B, MNDA, SERPINB4, CLEC3B, PLEC, LGALS7, H2AC4, SERPIN B3, AMBP, PFN1, DSC3, IGHG1, ORM1, RECQL, RPL26, KLK13, RPL22, APOA2, SERPINB5, At least one protein selected from the group consisting of LCN15, IGHG3, CAP1, and SPRR2F; , a group consisting of proteins shown in Tables 11-1 to 11-4, Tables 12-1 to 12-4 and Table 13 In combination with at least one other protein selected from The method described in [1] or [2]. [5] The subject is preferably an adult, the at least one protein Preferably, at least one protein selected from the group consisting of proteins shown in Tables 5-1 to 5-9. is also a type of protein, More preferably, from the proteins shown in Tables 9-1 to 9-7 and Tables 10-1 to 10-2 at least one protein selected from the group consisting of More preferably, the protein is selected from the group consisting of proteins shown in Tables 14-1 to 14-7. or at least one protein listed in Tables 15-1 to 15-4. or at least one protein selected from the group consisting of a protein shown in Table 16. at least one protein selected from the group consisting of proteins, More preferably, SERPINB1, TTR, DHX36, ITIH4, GC, ALB, SERPING1, DDX55, and IGH Selected from the group consisting of V1-46, EZR, VTN, AHSG, HPX, PPIA, KNG1, FN1, PLG, PRDX6 and FLG2 at least one protein selected from the group consisting of More preferably, SERPINB1, TTR, DHX36, ITIH4, GC, ALB, SERPING1, DDX55, and IGH Selected from the group consisting of V1-46, EZR, VTN, AHSG, HPX, PPIA, KNG1, FN1, PLG, PRDX6 and FLG2 At least one protein selected from Tables 14-1 to 14-7 and Tables 15-1 to 15-4 and at least one other protein selected from the group consisting of the proteins shown in Table 16. It is a combination of The method described in [1] or [2]. [6] The subject is preferably an infant, The at least one protein is preferably selected from the proteins shown in Tables 7-1 to 7-4. at least one protein selected from the group consisting of The method preferably comprises detecting an increase in the concentration of the at least one protein compared to a healthy group. and detecting the subject as having atopic dermatitis when the subject has [2] The method described in [2]. [7] The subject is preferably an infant, The at least one protein is preferably selected from the group consisting of the proteins shown in Table 8. at least one protein selected from The method preferably comprises detecting a decrease in the concentration of the at least one protein compared to a healthy group. and detecting the subject as having atopic dermatitis when the subject has [2] The method described in [2]. [8] The subject is preferably an adult, The at least one protein is preferably selected from the proteins shown in Tables 9-1 to 9-7. at least one protein selected from the group consisting of The method preferably comprises detecting an increase in the concentration of the at least one protein compared to a healthy group. and detecting the subject as having atopic dermatitis when the subject has [2] The method described in [2]. [9] The subject is preferably an adult, The at least one protein is preferably a protein shown in Tables 10-1 to 10-2. at least one protein selected from the group consisting of proteins, The method preferably comprises detecting a decrease in the concentration of the at least one protein compared to a healthy group. and detecting the subject as having atopic dermatitis when the subject has [2] The method described in [2].

[10] The method comprises: Preferably, the concentration of the at least one protein is used as an explanatory variable to determine whether or not AD is present. detecting AD based on a predictive model constructed as a functional variable; More preferably, A is determined based on a cutoff value for discriminating between atopic dermatitis patients and healthy individuals. D, wherein the cutoff value is set to the at least one of the above-mentioned The concentration of another protein and the concentration of the protein derived from a healthy subject were used as teacher samples. The skin of the subject is calculated from a discriminant formula for distinguishing between atopic dermatitis patients and healthy individuals. The concentration of at least one protein obtained from the lipids on the skin surface is substituted into the discriminant. The result is compared with the cutoff value to determine the atopic dermatitis in the subject. Evaluate the presence or absence of The method according to any one of [2] to [5].

[11] Preferably, a skin sample from a subject having or suspected of developing atopic dermatitis. The method according to any one of [2] to

[10] , wherein lipids on the skin surface are detected.

[12] The subject is preferably an infant, The at least one protein is preferably selected from the proteins shown in Tables 7-1 to 7-4. at least one protein selected from the group consisting of The method preferably comprises detecting an increase in the concentration of the at least one protein compared to a healthy group. If the lipids on the skin surface are present, the person has atopic dermatitis or is suspected of developing atopic dermatitis. detecting the antibody as originating from the subject.

[11] The method described in.

[13] The subject is preferably an infant, The at least one protein is preferably selected from the group consisting of the proteins shown in Table 8. at least one protein selected from The method preferably comprises detecting a decrease in the concentration of the at least one protein compared to a healthy group. If the lipids on the skin surface are present, the person has atopic dermatitis or is suspected of developing atopic dermatitis. detecting the antibody as originating from the subject.

[11] The method described in.

[14] The subject is preferably an adult, The at least one protein is preferably selected from the proteins shown in Tables 9-1 to 9-7. at least one protein selected from the group consisting of The method preferably comprises detecting an increase in the concentration of the at least one protein compared to a healthy group. If the lipids on the skin surface are present, the person has atopic dermatitis or is suspected of developing atopic dermatitis. detecting the antibody as originating from the subject.

[11] The method described in.

[15] The subject is preferably an adult, The at least one protein is preferably a protein shown in Tables 10-1 to 10-2. at least one protein selected from the group consisting of proteins, The method preferably comprises detecting a decrease in the concentration of the at least one protein compared to a healthy group. If the lipids on the skin surface are present, the person has atopic dermatitis or is suspected of developing atopic dermatitis. detecting the antibody as originating from the subject.

[11] The method described in.

[16] Preferably, the method further comprises collecting lipids on the skin surface from the subject. The method described in any one of

[15] to

[15] .

[0095]

[17] At least one protein selected from the group consisting of proteins shown in Tables 1-1 to 1-13. Protein markers for detecting atopic dermatitis, including one type.

[18] Preferably, the at least one protein is: At least one protein selected from the group consisting of proteins shown in Tables 2-1 to 2-5 Is it high quality? At least one protein selected from the group consisting of proteins shown in Tables 3-1 to 3-2 It is of high quality,

[17] The marker described in

[17] .

[19] The marker is preferably a marker for detecting atopic dermatitis in infants, the at least one protein Preferably, the protein is selected from the group consisting of proteins shown in Tables 7-1 to 7-4 and Table 8. at least one protein, More preferably, it is selected from the group consisting of proteins shown in Tables 11-1 to 11-4. at least one protein, More preferably, a small amount of a protein selected from the group consisting of proteins shown in Tables 4-1 to 4-6. At least one protein, More preferably, POF1B, MNDA, SERPINB4, CLEC3B, PLEC, LGALS7, H2AC4, SERPIN B3, AMBP, PFN1, DSC3, IGHG1, ORM1, RECQL, RPL26, KLK13, RPL22, APOA2, SERPINB5, At least one protein selected from the group consisting of LCN15, IGHG3, CAP1, and SPRR2F be,

[17] The marker described in

[17] .

[20] The subject is preferably a subject, and the marker is preferably a marker for detecting atopic dermatitis in adults. It is a marker, the at least one protein Preferably, the protein is one shown in Tables 9-1 to 9-7 and 10-1 to 10-2. at least one protein selected from the group More preferably, it is selected from the group consisting of the proteins shown in Tables 14-1 to 14-7. at least one protein, More preferably, a small amount of a protein selected from the group consisting of proteins shown in Tables 5-1 to 5-9. At least one protein, More preferably, SERPINB1, TTR, DHX36, ITIH4, GC, ALB, SERPING1, DDX55, and IGH Selected from the group consisting of V1-46, EZR, VTN, AHSG, HPX, PPIA, KNG1, FN1, PLG, PRDX6 and FLG2 at least one protein selected from the group consisting of

[17] The marker described in

[17] .

[0096]

[21] At least one protein selected from the group consisting of proteins shown in Tables 1-1 to 1-13 above. Use of a single protein as a marker for detecting atopic dermatitis.

[22] At least one protein selected from the group consisting of proteins shown in Tables 1-1 to 1-13 above. Use of a protein in the production of a protein marker for detecting atopic dermatitis.

[23] Preferably, the at least one protein is: At least one protein selected from the group consisting of proteins shown in Tables 2-1 to 2-5 Is it high quality? At least one protein selected from the group consisting of proteins shown in Tables 3-1 to 3-2 It is of high quality, The use described in

[21] or

[22] .

[24] The marker is preferably a marker for detecting atopic dermatitis in infants, the at least one protein Preferably, the protein is selected from the group consisting of proteins shown in Tables 7-1 to 7-4 and Table 8. at least one protein, More preferably, it is selected from the group consisting of proteins shown in Tables 11-1 to 11-4. at least one protein, More preferably, a small amount of a protein selected from the group consisting of proteins shown in Tables 4-1 to 4-6. At least one protein, More preferably, POF1B, MNDA, SERPINB4, CLEC3B, PLEC, LGALS7, H2AC4, SERPIN B3, AMBP, PFN1, DSC3, IGHG1, ORM1, RECQL, RPL26, KLK13, RPL22, APOA2, SERPINB5, At least one protein selected from the group consisting of LCN15, IGHG3, CAP1, and SPRR2F be, The use described in

[21] or

[22] .

[25] The subject is preferably a subject, and the marker is preferably a marker for detecting atopic dermatitis in adults. It is a marker, the at least one protein Preferably, the protein is one shown in Tables 9-1 to 9-7 and 10-1 to 10-2. at least one protein selected from the group More preferably, it is selected from the group consisting of the proteins shown in Tables 14-1 to 14-7. at least one protein, More preferably, a small amount of a protein selected from the group consisting of proteins shown in Tables 5-1 to 5-9. At least one protein, More preferably, SERPINB1, TTR, DHX36, ITIH4, GC, ALB, SERPING1, DDX55, and IGH Selected from the group consisting of V1-46, EZR, VTN, AHSG, HPX, PPIA, KNG1, FN1, PLG, PRDX6 and FLG2 at least one protein selected from the group consisting of The use described in

[21] or

[22] . [Example]

[0097] The present invention will be described in more detail below with reference to examples, but the present invention is not limited to these examples. It's not that.

[0098] Example 1: Analysis of atopic dermatitis-related expression changes using proteins derived from infant SSL Protein identification 1) Subjects and SSL collection 23 healthy children (6 months to 5 years old, both male and female) (healthy group) and 23 children with atopic dermatitis (AD group) The subjects were 16 children (6 months to 5 years old, both male and female) (AD group). We collected children with AD who met the UKWP criteria as assessed by their parents and administered an informed consent. We selected patients whose guardians provided consent through a licensed physician. A dermatologist will observe and interview the skin of the whole body, and will determine the cause of the atopic dermatitis based on the guidelines for its treatment. Among the children diagnosed with AD, atopic dermatitis treatment guide was Based on the severity assessment criteria described in the guidelines, the patient was diagnosed with mild to severe AD-like eczema or dryness on the face. Infants and toddlers exhibiting symptoms were selected as subjects. Each subject's entire face (including the rash area in the case of children with AD) was examined. Remove sebum from the skin using an oil absorbing film (5 x 8 cm, polypropylene, 3M). The blotting film was transferred to a glass vial and stored until ready for protein extraction. The mixture was stored at −80°C for approximately one month.

[0099] 2) Protein preparation Cut the oil absorbing film from 1) above to an appropriate size and place it in QIAzol Lysis Protein precipitation was performed using Qiagen Reagent according to the attached protocol. From the resulting protein precipitate, MPEX PTS Reagent (GL s The protein was dissolved in the solubilization solution according to the attached protocol using a centrifuge tube. The peptide solution was then digested with lipase to obtain a peptide solution. The resulting peptide solution was dried under reduced pressure (35°C). Then, add water containing 0.1% formic acid and 2% acetonitrile. The solution was dissolved in Pierce. TM Quantitative Fluoromet ric Peptide Assay(ThermoFisher Scientifi Following the protocol in c), the plate was read using a microplate reader (Corona Electric The peptide concentration in the solution was measured using the ELISA kit. The peptide solution from one patient was excluded from the samples analyzed below. The peptide concentration fed to the MS instrument was kept constant during the analysis, and the protein quantification value was was calculated.

[0100] 3) LC-MS / MS analysis and data analysis The sample peptide solution obtained in 2) above was analyzed by LC-MS / MS under the conditions in Table 6 below. was served.

[0101] [Table 6]

[0102] The spectral data obtained by LC-MS / MS analysis is analyzed using proteome analysis. Discoverer ver.2.2(ThermoFisher Scientif For protein identification, the reference databases were Swiss Prot, Tax Set the onomy to Homo sapiens and use the Mascot database s The search was performed using the Enzyme search (Matrix Science). e is Trypsin, Missed cleavage is 2, Dynamic mo Oxidation (M), Acetyl (N-term), Acetyl (Protein N-term), Static Modificati ons was set to Carbamidomethyl(C). False disco Peptides satisfying the FDR p<0.01 were searched for. Label-free quantitative analysis of the proteins using precursor ions (LF) Q, Label-Free Quantification was performed. Protein quantification values ​​are calculated based on the peak intensity of the precursor ion, and the peak intensity is detected. Values ​​below the threshold were considered missing. The variance-based method was used to calculate the p-value, which indicates the significance of the difference between groups. ANOVA (Individual Based, t-test) was used.

[0103] 4) Results Among the identified proteins, the false discovery rate (FDR) Proteins with a value of 0.1 or higher were excluded from the analysis. 533 proteins for which non-missing protein values ​​were calculated in 75% or more of the subjects were included. Proteins with a prevalence of 1.5 times or more (p<0.01) in the AD group compared to the healthy group were extracted for analysis. 116 proteins (Tables 7-1 to 7-4) whose levels increased by 0.75 times or more (≤0.05), and Twelve proteins (Table 8) that were down-regulated (p≦0.05) were identified.

[0104] [Table 7-1]

[0105] [Table 7-2]

[0106] [Table 7-3]

[0107] [Table 7-4]

[0108] [Table 8]

[0109] Example 2: Analysis of differentially expressed proteins associated with atopic dermatitis using adult SSL-derived proteins Identification of substances 1) Subjects and SSL collection 18 healthy individuals (20-59 years old, male) (healthy group) and 18 patients with atopic dermatitis (AD patients) The subjects were 26 men (20-59 years old) (AD group). Informed consent was obtained. The subjects in the AD group were assessed for severity by a dermatologist. Those who have been diagnosed with mild or moderate atopic dermatitis and have mild to severe AD-like eczema on the face The subjects were selected from those who presented with symptoms such as dryness and irritation. Sebum was collected using an oil blotting film (5 x 8 cm, polypropylene, 3M). The blotting film was transferred to a vial and stored at -80°C until used for protein extraction. It was stored for about a month.

[0110] 2) Protein preparation MPEX PTS Reagent (GL science) replaced with EasyPep TM Mini MS Sample Prep Kit(ThermoFisher S Scientific) and obtained the peptide solution according to the attached protocol. The peptide concentration was measured in the same manner as in Example 1.

[0111] 3) LC-MS / MS analysis and data analysis Protein analysis and data analysis were carried out under the same conditions and procedures as in Example 1.

[0112] 4) Results Among the identified proteins, the false discovery rate (FDR) Proteins with a value of 0.1 or higher were excluded from the analysis. 1075 proteins for which non-missing protein values ​​were calculated in 75% or more of the subjects Proteins were extracted as the analysis target. One patient with D was excluded from the analysis. 205 proteins (Tables 9-1 to 9-7) whose levels increased by 0.75 times or more (≤0.05) and We identified 37 proteins (Tables 10-1 to 10-2) that were significantly decreased (p≦0.05). .

[0113] [Table 9-1]

[0114] [Table 9-2]

[0115] [Table 9-3]

[0116] [Table 9-4]

[0117] [Table 9-5]

[0118] [Table 9-6]

[0119] [Table 9-7]

[0120] [Table 10-1]

[0121] [Table 10-2]

[0122] Example 3: Construction of a discriminant model for detecting atopic dermatitis in infants and young children (Usage Data) In order to approximate the quantitative data of proteins obtained in Example 1 to a normal distribution, The peak intensity before the rise was used as the protein quantification value, and each protein quantification value was used as the total detected protein. The value divided by the sum of the quantitative values ​​was converted to a logarithmic value of base 2, Log2(Abundance The Log2(Abundance+1) value was calculated. The obtained Log2(Abundance+1) value was used as the machine learning model. In the same manner as in Example 1, defects were detected in more than 75% of all subjects (more than 29 subjects). 475 proteins for which quantitative values ​​other than the standard deviation were calculated were extracted and analyzed. did.

[0123] 3-1 Construction of a discrimination model using differentially expressed proteins 1) Selection of feature proteins Among the 475 proteins listed above, those found to be statistically significantly higher in children with AD than in healthy children We identified 127 proteins whose expression levels were altered (Tables 11-1 to 11-4). The proteins were selected as feature proteins, and their quantitative data were used as features. 2) Model construction The Log2(Abundance+1) values ​​of the above 127 proteins were used as explanatory variables, The objective variables were healthy children and children with AD (whether or not AD was present). In the package, the random forest algorithm is specified as the method, and a single decision is made. The number of variables used to construct the tree (mtry value) was tuned to the optimum value. Run the random forest algorithm using the mtry values ​​determined by The OB error rate was calculated. As a result, 127 proteins were identified as feature proteins. When used as a protein, the model had an error rate of 18.42%.

[0124] [Table 11-1]

[0125] [Table 11-2]

[0126] [Table 11-3]

[0127] [Table 11-4]

[0128] 3-2 Construction of a discriminant model using proteins with high variable importance in random forests 1) Selection of feature proteins The Log2(Abundance+1) values ​​of the above 475 proteins were used as explanatory variables. The objective variables were healthy children and children with AD (whether or not AD was present). In the package, specify the random forest algorithm as the method and run a single The optimal value of the number of variables (mtry value) used to construct the decision tree was tuned. Using the mtry values ​​determined by the algorithm, a random forest algorithm is run. The top 140 proteins with variable importance based on the Gini coefficient were calculated (Tables 12-1 to 12-4 ) These 140 proteins and all 475 proteins used to select feature proteins Proteins were used as feature proteins, and their quantitative data were used as features. 2) Model construction The Log2 (Abundance) of the above 140 proteins or all 475 proteins e+1) was used as the explanatory variable, and healthy children and children with AD (presence or absence of AD) were used as the dependent variable. The random forest algorithm is implemented in the R package "caret". The optimal number of variables (mtry value) used to build a decision tree is determined by the Using the mtry value determined by tuning, a random forest was We ran the algorithm and calculated the estimated OOB error rate. As a result, the error rate when using all 475 proteins as feature proteins was 28.95%. On the other hand, when the top 140 proteins in terms of variable importance were used as feature proteins, The error rate for all cases was 7.89%.

[0129] [Table 12-1]

[0130] [Table 12-2]

[0131] [Table 12-3]

[0132] [Table 12-4]

[0133] 3-3 Construction of a discriminant model using feature proteins extracted by the Boruta method 1) Selection of feature proteins The Log2(Abundance+1) values ​​of the above 475 proteins were used as explanatory variables. Using healthy children and children with AD (whether or not AD was present) as the objective variables, the data was analyzed using the R program "Boruta". The algorithm in the package was run with a maximum of 1000 trials and a p-value of 0.01. Thirty-five proteins with a molecular weight of less than 1000 were extracted (Table 13) and selected as feature proteins. The quantitative data of these proteins was used as features. 2) Model construction The Log2(Abundance+1) values ​​of the above 35 proteins were used as explanatory variables. The objective variables were normal children and children with AD (whether or not AD was present). In the cage, the random forest algorithm is specified as the method, and a single decision is made. The optimal value of the number of variables (mtry value) used to construct the tree was tuned. Using the determined mtry value, the random forest algorithm is run. As a result, 35 proteins were identified as feature proteins. When used as quality, the model had an error rate of 10.53%.

[0134] [Table 13]

[0135] Example 4: Construction of a discriminant model for detecting atopic dermatitis in adults (Usage Data) In order to approximate the quantitative data of proteins obtained in Example 2 to a normal distribution, The peak intensity before the rise was used as the protein quantification value, and each protein quantification value was used as the total detected protein. The value divided by the sum of the quantitative values ​​was converted to a logarithmic value of base 2, Log2(Abundance The Log2(Abundance+1) value was calculated. The obtained Log2(Abundance+1) value was used as the machine learning model. The same method as in Example 2 was used to construct the genomic DNA of all subjects (excluding those for whom quantitative data on proteins was available). Proteins with non-missing values ​​in 75% or more (31 or more people) of the total (excluding 3 people who do not follow a normal distribution) 985 types of proteins whose quantitative values ​​were calculated were extracted and analyzed.

[0136] 4-1 Construction of a discrimination model using differentially expressed proteins 1) Selection of feature proteins Among the above 985 proteins, the expression of which was significantly altered in AD patients compared to healthy individuals was The 220 proteins (Tables 14-1 to 14-7) were identified and analyzed as feature proteins. The quantitative data was used as the feature. 2) Model construction The Log2(Abundance+1) values ​​of the above 220 proteins were used as explanatory variables, Healthy individuals and AD patients (whether or not AD was present) were used as the objective variables. In the cage, the random forest algorithm is specified as the method, and a single decision is made. The optimal value of the number of variables (mtry value) used to construct the tree was tuned. Using the determined mtry value, the random forest algorithm is run. As a result, 220 proteins were identified as feature proteins. When used as quality, the model had an error rate of 24.39%.

[0137] [Table 14-1]

[0138] [Table 14-2]

[0139] [Table 14-3]

[0140] [Table 14-4]

[0141] [Table 14-5]

[0142] [Table 14-6]

[0143] [Table 14-7]

[0144] 4-2 Construction of a discriminant model using proteins with high variable importance in random forests 1) Selection of feature proteins The Log2(Abundance+1) values ​​of the above 985 proteins were used as explanatory variables. The target variables were healthy subjects and AD patients (whether or not AD was present). In the package, the random forest algorithm is specified as the method, and a single decision is made. The number of variables used to construct the tree (mtry value) was tuned to the optimum value. Using the mtry value determined by the above, the random forest algorithm was run and the The top 110 proteins with variable importance based on the Ni coefficient were calculated (Tables 15-1 to 15-4). These 110 proteins and all 985 proteins used to select feature proteins Proteins were used as feature proteins, and their quantitative data were used as features. 2) Model construction The Log2 (Abundance) of the above 110 proteins or all 985 proteins e+1) value was used as the explanatory variable, and healthy subjects and AD patients (with or without AD) were used as the dependent variable. The random forest algorithm is implemented in the R language "caret" package. The optimal number of variables (mtry value) used to build a decision tree is selected as the The mtry value determined by tuning was used to calculate the random forest The algorithm was run and the estimated OOB error rate was calculated. As a result, the error rate when all 985 proteins were used as feature proteins was 29.27%. In contrast, when the top 110 proteins in terms of variable importance were used as feature proteins, The error rate was 12.20%.

[0145] [Table 15-1]

[0146] [Table 15-2]

[0147] [Table 15-3]

[0148] [Table 15-4]

[0149] 4-3 Construction of a discriminant model using features extracted by the Boruta method 1) Feature selection The Log2(Abundance+1) values ​​of the above 985 proteins were used as explanatory variables. Using healthy subjects and AD patients (whether or not AD was present) as the objective variables, the “Boruta” program in R was used. The algorithm in the package was run. The maximum number of trials was 1000, and the p-value was less than 0.01. 24 proteins were extracted (Table 16) and selected as feature proteins. The quantitative data of these proteins were used as features. 2) Model construction The Log2(Abundance+1) values ​​of the above 24 proteins were used as explanatory variables. The target variables were normal subjects and AD patients (whether or not AD was present). Specify the random forest algorithm as the method in the page and run a single decision tree. The optimal value of the number of variables (mtry value) used to construct the Using the determined mtry value, the random forest algorithm is run and OOB The error rate was calculated. As a result, 24 proteins were identified as feature proteins. The model using as had an error rate of 19.51%.

[0150] [Table 16]

[0151] A total of 418 proteins (Tables 1-1 to 1-13) were obtained through the analysis of the above examples. ) in a paper reporting its association with AD through text mining (Elsevier) The search revealed that there were four or fewer reports related to AD and that the relationship with AD was not described. There were 147 proteins that were confirmed not to be present (Tables 2-1 to 2-5 above). These 147 proteins are novel markers for detecting AD.

Claims

[Claim 1] 1. A method for measuring a protein expression level for detecting atopic dermatitis in a subject, comprising: Measuring the expression level of one protein selected from the group consisting of proteins shown in Table 1 below from lipids on the skin surface collected from the entire face of a subject; detecting the lipids on the skin surface as being derived from a subject with atopic dermatitis when the expression level of the one protein measured is increased compared to a reference value for distinguishing between atopic dermatitis patients and healthy individuals; Including, the subject is an infant, method. Table 1

Images