Construction method of diagnosis model of depression, detection method of depression and application

Abstract

The invention provides a construction method of a diagnosis model of depression, a detection method of depression and application. Wherein the diagnosis model is used for diagnosing whether women in the non-perinatal period suffer from depression or not; the construction method comprises the following steps: a) acquiring depression conditions and biological samples of a sample population, detecting biomarkers in the biological samples, and establishing a data set of the biological samples, the data set comprises a mapping relationship between the depression condition of the sample population and the content of the biomarker in the biological sample; b) utilizing a machine learning method to construct the diagnosis model, the sample crowds comprise healthy crowds in non-perinatal women and depression patient crowds, and the biomarkers comprise 25-OH VD. The problem of low diagnosis accuracy of female depression in the prior art can be solved, and the method is suitable for the field of depression detection.

Description

Technical Field

[0001] This invention relates to the field of depression detection, and more specifically, to a method for constructing a diagnostic model for depression, a method for detecting depression, and its application. Background Technology

[0002] Depression, or Major Depressive Disorder (MDD), is a common and serious mental health disorder characterized by single or recurrent episodes with a high risk of relapse. Clinical diagnosis typically describes it as a depressive episode or recurrent depressive disorder. Clinical observations indicate that approximately 20-30% of patients with depression experience a long-term, chronic course with a poor prognosis, known as chronic depressive disorder. This further highlights the complexity and difficulty of managing depression.

[0003] Clinically, patients with depression often exhibit core symptoms such as depressed mood, grief, and stupor. However, these symptoms are not consistent across all patients; some may also experience anxiety, agitation, and in more severe cases, hallucinations, delusions, self-harm, or even suicidal tendencies may occur. This diversity and complexity of symptoms presents challenges to the clinical diagnosis of depression.

[0004] Despite extensive research, a consensus on the exact cause of MDD remains elusive. The neuroendocrine imbalance hypothesis proposes that functional imbalances and homeostasis disturbances of neurotransmitters such as serotonin (5-HT), dopamine (DA), and norepinephrine (NE) may be key to the pathogenesis of depression. The neuroimmunological hypothesis explores the interaction between the nervous and immune systems; although the exact mechanisms are unclear, studies have shown a significantly increased incidence of depression in patients receiving immunotherapy and those with autoimmune diseases. Currently, research into the pathogenesis of depression has driven the development of objective diagnostic biomarkers, involving markers related to mental impairment, endocrine hormones, and humoral homeostasis, as well as markers related to various metabolic pathways such as lipid metabolism and glucose metabolism. Some literature also reports a certain correlation with neuroimmunology. Although electrophysiological and imaging studies attempt to explore the pathology of depression from the perspective of brain structure and function, and various pathogenic hypotheses exist, consistent research results have yet to be obtained. No diagnostic biomarkers or models have been reported that can be universally applied across populations, limiting their clinical translation and application.

[0005] Currently, the diagnosis of depression is mainly based on the DSM-5 criteria, with clinical diagnosis through psychiatric interviews and the Hamilton Depression Rating Scale (HAMD-24) used to assess the severity of depression. However, the diagnostic accuracy of relying primarily on clinical manifestations and scales for the diagnosis of depression is not high, and nearly four-fifths of patients with depressive disorders are misdiagnosed or missed.

[0006] Furthermore, research shows that depressive disorders exhibit significant gender and age-related heterogeneity, with the global incidence rate in women being approximately 1.5 to 2 times that of men. This significant difference is closely related to the dynamic changes in women's unique physiological stages: the rapid rise in estrogen levels during puberty significantly increases the vulnerability of mood regulation by affecting hippocampal neurogenesis and serotonin receptor sensitivity; fluctuations in thyroid hormones during reproductive years interact with the disruption of cortisol's diurnal rhythm, exacerbating susceptibility to neuroendocrine disorders; and perimenopause further amplifies the risk of mood regulation imbalance due to decreased metabolic rate and altered immune response sensitivity. However, current research on biomarkers for depressive disorders is mostly based on mixed-sex or broad age range samples, without targeted calibration for the characteristics of hormonal fluctuations, differences in metabolic patterns, and specificity of immune responses at different physiological stages in women. Additionally, there is no clear guidance on which of the many pathogenic hypotheses are more relevant to the pathogenesis in women, leading to a decline in the diagnostic sensitivity and specificity of existing biomarkers in the female population. Therefore, for the high-incidence and highly heterogeneous group of women aged 19-65, constructing diagnostic criteria or models for depression that are independent of male samples has important clinical value for achieving more accurate disease identification. Summary of the Invention

[0007] The main objective of this invention is to provide a method for constructing a diagnostic model for depression, a method for detecting depression, and its application, in order to solve the problem of low diagnostic accuracy for depression in women in the prior art.

[0008] To achieve the above objectives, according to a first aspect of the present invention, a method for constructing a diagnostic model for depression is provided, the diagnostic model being used to diagnose whether non-perinatal women suffer from depression; the method comprising: a) acquiring depression status and biological samples from a sample population, detecting biomarkers in the biological samples, and establishing a dataset of the biological samples, the dataset including a mapping relationship between depression status in the sample population and the content of the biomarkers in the biological samples; b) constructing the diagnostic model using machine learning methods; the sample population including healthy individuals and patients with depression among non-perinatal women; wherein the biomarker includes 25-OH VD.

[0009] Further, the above-mentioned biomarkers also include one or more of triiodothyronine, cortisol, TNFα, or IL-6; optionally, the above-mentioned biomarkers include any of the following combinations: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH VD + IL-6; 6) 25-OH VD + IL-6 + TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + cortisol + IL-6; 12) 25-OH VD + cortisol + IL-6 + TNFα; 13) 25-OH VD + triiodothyronine + IL-6 + TNFα; VD+triiodothyronine+cortisol+TNFα; 14)25-OH VD+triiodothyronine+cortisol+IL-6; 15)25-OHVD+triiodothyronine+cortisol+IL-6+TNFα; Preferably, the above-mentioned triiodothyronine is free triiodothyronine.

[0010] Furthermore, the aforementioned machine learning methods include any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost; preferably, before constructing the aforementioned diagnostic model using any one of the aforementioned logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, and Xgboost, the construction method further includes: dividing the aforementioned dataset into a training set and a test set, constructing a classification model using the aforementioned training set, and verifying the model's performance using the aforementioned test set.

[0011] Furthermore, the aforementioned non-perinatal women are non-perinatal women aged 19-65; preferably, the aforementioned biological sample is one of serum, plasma, or whole blood sample.

[0012] To achieve the above objectives, according to a second aspect of the present invention, a method for detecting depression is provided. This method is used to diagnose whether a subject suffers from depression or to detect the risk of depression in a subject, wherein the subject is a non-perinatal woman. The method includes: i) collecting biological samples from the subject, detecting the content of biomarkers in the method for constructing a diagnostic model for depression in the biological samples, and obtaining characteristic data of the biomarkers in the subject's biological samples; ii) generating a diagnosis result for depression or a prediction result for the risk of depressive disorder based on the characteristic data of the biomarkers in the subject's biological samples; optionally, the characteristic data of the biomarkers in the subject's biological samples are selected from the content of the biomarkers or data after standardization of the content of the biomarkers; optionally, ii) includes: ii-1) comparing the content of the biomarkers in the subject with a first predetermined threshold, and generating a diagnosis result for depression or a prediction result for the risk of depression based on the comparison result; or ii-2) [The text abruptly ends here, so the translation stops.] The standardized data is input into a pre-established depression diagnostic model, and the calculated result is compared with a second predetermined threshold. Based on the comparison result, a depression diagnosis result or a depression risk prediction result for the subject is generated. Preferably, the first predetermined threshold is a threshold obtained based on an ROC curve generated directly without a machine learning model, or a threshold obtained based on an ROC curve generated by a machine learning model. Preferably, the second predetermined threshold is a predicted probability threshold or a dependent variable threshold of the pre-established diagnostic model. Optionally, the diagnostic model is constructed by machine learning based on known depression diagnoses of non-perinatal women and the content of the biomarkers in biological samples of non-perinatal women. Preferably, the machine learning includes any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost. Preferably, the diagnostic model is a diagnostic model obtained using the above-mentioned method for constructing a depression diagnostic model.

[0013] To achieve the above objectives, according to a third aspect of the present invention, a detection reagent is provided for use in the preparation of a product for diagnosing depression in non-perinatal women, the detection reagent being capable of detecting the content of biomarkers in biological samples from non-perinatal women, the biomarkers including 25-OH VD.

[0014] Further, the above-mentioned biomarkers also include one or more of triiodothyronine, cortisol, TNFα, or IL-6; optionally, the above-mentioned biomarkers include any of the following combinations: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH VD + IL-6; 6) 25-OH VD + IL-6 + TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + cortisol + IL-6; 12) 25-OH VD + cortisol + IL-6 + TNFα; 13) 25-OH VD + triiodothyronine + IL-6 + TNFα; VD+triiodothyronine+cortisol+TNFα; 14)25-OH VD+triiodothyronine+cortisol+IL-6; 15)25-OHVD+triiodothyronine+cortisol+IL-6+TNFα.

[0015] Furthermore, the detection method for the content of the above-mentioned biomarkers includes one or more of enzyme-linked immunosorbent assay (ELISA), chemiluminescent immunoassay, serum biochemical assay, high performance liquid chromatography (HPLC), or mass spectrometry (MS). Preferably, the above-mentioned product includes a kit or electronic predictive device. Preferably, the above-mentioned detection reagent includes any one or more of the following: an antibody capable of binding to the above-mentioned biomarker, a tracer for labeling the above-mentioned antibody, a standard of the above-mentioned biomarker, a solid-phase carrier for antibody-antigen binding reaction, or a buffer solution.

[0016] To achieve the above objectives, according to a fourth aspect of the present invention, an electronic device is provided for diagnosing whether a non-perinatal adult woman suffers from depression, or for detecting the risk of depression in a non-perinatal adult woman. The electronic device includes a data collection module and a diagnostic module. The data collection module is used to collect the levels of biomarkers in a biological sample of a subject and input the data collected by the data collection module into the diagnostic module. The biomarkers include 25-OH. VD; The diagnostic module is configured to output the diagnostic result or risk prediction result of the subject based on the biomarker content in the subject's biological sample; Optionally, the diagnostic module contains a diagnostic model, which calculates and outputs the judgment result using the data collected by the data collection module and the diagnostic model; The diagnostic model is constructed based on the known depression diagnosis results of non-perinatal women and the biomarker content in the biological sample of the non-perinatal women using any one of the following methods: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost; Optionally, the biomarker also includes one or more of triiodothyronine, cortisol, TNFα, or IL-6; Optionally, the biomarker includes any one of the following combinations: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH 6) 25-OH VD+IL-6+TNFα; 7) 25-OH VD+Triiodothyronine+TNFα; 8) 25-OH VD+Triiodothyronine+IL-6; 9) 25-OH VD+Triiodothyronine+IL-6+TNFα; 10) 25-OH VD+Cortisol+TNFα; 11) 25-OH VD+Cortisol+IL-6; 12) 25-OH VD+Cortisol+IL-6+TNFα; 13) 25-OH VD+Triiodothyronine+Cortisol+TNFα; 14) 25-OH VD+Triiodothyronine+Cortisol+IL-6; 15) 25-OH VD+triiodothyronine+cortisol+IL-6+TNFα; Optionally, the diagnostic module uses the content of the above biomarkers and the above diagnostic model to calculate and output the judgment result, including: inputting the content of the above biomarkers into the above diagnostic model after standardization to obtain a target value; comparing the above target value with a predetermined threshold to obtain the diagnosis result or risk prediction result of depression in the above non-perinatal women; preferably, the above diagnostic model is a diagnostic model obtained using the above method for constructing a diagnostic model for depression.

[0017] To achieve the above objectives, according to a fifth aspect of the present invention, a method for detecting depression is provided. The steps of the method are performed by a computer. The method is used to diagnose whether a subject suffers from depression or to detect the risk of depression in a subject who is a non-perinatal woman. The method includes: generating a depression risk prediction result for the subject based on characteristic data of biomarkers in the subject's biological sample; optionally, the characteristic data of the biomarkers in the subject's biological sample is selected from the content of the biomarkers or data after standardization of the biomarker content; optionally, the method includes: comparing the content of the subject's biomarkers with a first predetermined threshold, and generating a depression diagnosis result or a depression risk prediction result for the subject based on the comparison result; or inputting the standardized data of the subject's biomarker content into a pre-established depression diagnosis model, and comparing the calculated result with a second predetermined threshold. The values ​​are compared, and based on the comparison results, a diagnosis result of depression or a prediction result of depression risk for the above-mentioned subjects is generated; preferably, the first predetermined threshold is a threshold obtained based on an ROC curve generated directly without a machine learning model, or a threshold obtained based on an ROC curve generated by a machine learning model; preferably, the second predetermined threshold is a predicted probability threshold or a dependent variable threshold of a pre-established diagnostic model; the diagnostic model is constructed by machine learning based on known depression diagnoses of non-perinatal women and the content of the above-mentioned biomarkers in biological samples of the above-mentioned non-perinatal women; preferably, the machine learning includes any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost; preferably, the diagnostic model is a diagnostic model obtained using the above-mentioned method for constructing a diagnostic model of depression.

[0018] To achieve the above objectives, according to a sixth aspect of the present invention, a kit is provided, comprising a detection reagent and instructions; the detection reagent comprises any one of the following reagents: A, a detection reagent for detecting 25-OH VD marker levels; B, a detection reagent for detecting 25-OH VD marker levels; and a detection reagent for detecting at least one marker level selected from triiodothyronine, Cortisol, IL6, or TNFα; the instructions indicate the use of the marker levels corresponding to the detection reagents for the diagnosis or risk prediction of depression in non-perinatal women.

[0019] To achieve the above objectives, according to a seventh aspect of the present invention, a computer-readable storage medium is provided, the storage medium including a stored program, wherein, when the program is executed, the device on which the storage medium is located is controlled to execute the method for constructing a diagnostic model for depression, or the method for detecting depression, or the method for detecting depression.

[0020] To achieve the above objectives, according to an eighth aspect of the present invention, a processor is provided for running a program, wherein the program executes the method for constructing a diagnostic model for depression, or the method for detecting depression, or the method for detecting depression.

[0021] By applying the technical solution of the present invention and utilizing the above-mentioned method for constructing a diagnostic model for depression, a diagnostic model applicable to diagnosing whether non-perinatal adult women suffer from depression can be obtained. This diagnostic model has higher accuracy in detecting the target population (non-perinatal adult women) than other general or targeted detection methods in the prior art. Attached Figure Description

[0022] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0023] Figure 1 The graph shows the results of a difference analysis of 25-OH VD between the depression group and the healthy group according to an embodiment of the present invention.

[0024] Figure 2 The ROC analysis results of 25-OH VD in distinguishing between the healthy group and the depressed group according to an embodiment of the present invention are shown. Detailed Implementation

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the embodiments.

[0026] As mentioned in the background section, existing technologies have developed some biomarkers and detection methods for the diagnosis of depression. However, the applicant found in actual research that the causes of depression are complex, and uniform detection methods have completely different detection effects in different populations, with significant group differences. Furthermore, existing technologies do not publicly show which method can effectively and accurately detect depression in non-perinatal adult women. Therefore, in this application, the inventors attempt to develop a method for diagnosing whether non-perinatal adult women suffer from depression, and based on this, propose a series of protection schemes for this application.

[0027] In a first typical embodiment of this application, a method for constructing a diagnostic model for depression is provided. The diagnostic model is used to diagnose whether non-perinatal women suffer from depression. The construction method includes: a) obtaining the depression status and biological samples of a sample population, detecting biomarkers in the biological samples, and establishing a dataset of biological samples, the dataset including the mapping relationship between the depression status of the sample population and the content of biomarkers in the biological samples; b) constructing a diagnostic model using machine learning methods; the sample population includes healthy individuals and patients with depression among non-perinatal women; wherein, the biomarker includes 25-OH VD.

[0028] This application proposes an innovative diagnostic prediction model construction method for the diagnosis of depression in non-perinatal adult women. Given that current diagnosis of depression primarily relies on subjective clinical assessment, resulting in high rates of misdiagnosis and missed diagnosis, we aim to improve the accuracy and efficiency of diagnosis by combining objective biomarker detection with advanced machine learning techniques. The following are more detailed implementation steps and explanations of the principles:

[0029] Step a) Dataset construction:

[0030] 1. Sample Collection: First, the sample population was selected, including a healthy control group of non-perinatal adult women and a group of women diagnosed with depression. This selection was based on ensuring the representativeness of the sample source so that the model could achieve the best predictive performance in the target population.

[0031] 2. Biological sample testing: Subsequently, the levels of biomarkers (including 25-OH VD) in the biological samples (such as blood, saliva, or urine) of each individual sample are thoroughly tested and analyzed.

[0032] 3. Data Mapping Establishment: The collected biological sample test data were paired with individual depression diagnoses to establish a comprehensive database including a healthy control group and a group of patients with depression. This database not only records the levels of various biomarkers, but more importantly, it forms a mapping relationship with individual depression states, laying the data foundation for subsequent predictive model construction.

[0033] Step b) Construction of the diagnostic prediction model:

[0034] 1. Selection and Application of Machine Learning Algorithms: Using the above dataset, supervised learning-based machine learning algorithms are employed for model training. Machine learning algorithms, including but not limited to logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machines, AdaBoost, DT decision trees, or XGBoost, are used to analyze the complex association between biomarkers and depressive states, thereby constructing a high-precision diagnostic prediction model.

[0035] Optionally, further feature engineering and model optimization can be performed: During model training, feature engineering operations are performed on various metrics in the dataset, including but not limited to data cleaning, missing value imputation, feature selection, and feature transformation, to improve the model's generalization ability and predictive performance. Model optimization uses techniques such as cross-validation and grid search to determine the optimal parameter configuration, ensuring the model's performance on unknown data.

[0036] 2. Model Evaluation and Validation: The completed model needs to undergo rigorous evaluation and validation to prove its effectiveness and stability. Optionally, one or more evaluation metrics, such as the area under the curve (AUC), accuracy, recall, and F1 score, can be used to comprehensively measure the model's classification ability. Furthermore, internal validation and external independent sample set testing are conducted to ensure the model's consistent performance across different data distributions.

[0037] In summary, the embodiments of this application aim to construct a diagnostic model capable of accurately assessing the depressive state of non-perinatal adult women by analyzing relevant data of biomarkers in biological samples and leveraging the power of machine learning. This method not only overcomes the limitations of traditional diagnostic methods, such as subjective bias and experience dependence, but also provides more objective and quantitative diagnostic results, facilitating early identification, stratified management, and the development of personalized treatment plans. Experimental results show that the model constructed using the aforementioned indicator components can accurately detect depression in the target population (i.e., non-perinatal adult women). Compared to other models or methods in the prior art, including but not limited to generalized methods for different populations or methods for other specific populations, the above-mentioned diagnostic prediction model has higher accuracy in detecting depression in the target population of non-perinatal adult women in this application.

[0038] 25-hydroxyvitamin D (abbreviated as 25-hydroxyVD or 25-OH VD): Vitamin D is a fat-soluble steroid derived from cholesterol, primarily composed of two forms: vitamin D2, also known as ergocalciferol, mainly obtained from vegetables and oral supplements; and vitamin D3, also known as cholecalciferol, mainly obtained through skin exposure to sunlight and from food (milk, fruit juice, fish, margarine, yogurt, grains, soybeans). Vitamin D is metabolized in the liver to 25-OH VD, which is the main circulating biomarker of vitamin D. In the kidneys, 25-OH VD is metabolized to 1,25-dihydroxyvitamin D by 25-OH VD-1α-hydroxylase. 25-OH VD is the most abundant circulating vitamin D metabolite and the best indicator of vitamin D status. Vitamin deficiency is a widespread problem in many populations worldwide; it is estimated that approximately 30% of children and 60% of adults worldwide are deficient or inadequate in vitamin D. People at risk of vitamin D deficiency typically include those with insufficient sun exposure, limited oral intake, or impaired intestinal absorption. Current methods for assessing vitamin D sufficiency primarily involve measuring the concentration of 25-OH VD in the blood.

[0039] In this application, 25-OH VD as a biomarker includes 25-OH VD in the peripheral blood of the subject.

[0040] In this application, the inventors discovered that a diagnostic model for depression can be constructed using 25-OH VD alone, and the constructed model is particularly suitable for the diagnosis of depression in non-perinatal women.

[0041] In a preferred embodiment, the biomarker further includes one or more of triiodothyronine, cortisol, TNFα, or IL-6; optionally, the biomarker includes any combination of the following: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH VD + IL-6; 6) 25-OH VD + IL-6 + TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + cortisol + IL-6; 12) 25-OH VD + cortisol + IL-6 + TNFα; 13) 25-OH VD + triiodothyronine + IL-6 + TNFα; VD+triiodothyronine+cortisol+TNFα; 14)25-OH VD+triiodothyronine+cortisol+IL-6; 15)25-OH VD+triiodothyronine+cortisol+IL-6+TNFα; Preferably, the triiodothyronine is free triiodothyronine.

[0042] Triiodothyronine (T3) is secreted by the thyroid gland, which produces and releases at least two hormones: total thyroxine (TT4) and total triiodothyronine (TT3). They play crucial roles in the human endocrine system, controlling metabolism, protein synthesis, carbohydrate and fat metabolism, neural development, normal bone growth and maturation, and cardiovascular and renal function. Approximately 99.7% of TT3 in circulation is bound to thyroid-binding globulin, with only 0.3% existing in a free form, FT3. Only these FT3 molecules, which can enter cells through specific membrane transport mechanisms, possess biological activity.

[0043] Cortisol is a glucocorticoid secreted by the adrenal pericortex and is an important biomarker of the adrenal glands. Corticotropin-releasing hormone (CRH) and adrenocorticotropic hormone (ACTH) stimulate cortisol secretion through feedback control. Cortisol can be detected in serum and urine. The main part of the cortisol circulation is bound to plasma corticosteroid-binding globulin (CBG or transcortisol) and albumin, preventing the hormone from penetrating the cell membrane of target cells. Only 3-5% of cortisol in plasma circulates in its unbound, free, biologically active form.

[0044] Tumor necrosis factor-alpha (TNF-α) is an inflammatory cytokine that plays a crucial role in apoptosis, cell survival, inflammation, and immunity. TNF-α is primarily produced by monocytes / macrophages and T cells, and is also expressed by neutrophils, NK cells, dendritic cells, endothelial cells, keratinocytes, astrocytes, and osteoblasts. TNF-α is initially synthesized as a 26 kDa transmembrane protein, which is then cleaved by the metalloproteinase TACE to form a soluble mature protein with a molecular weight of approximately 17 kDa. TNF-α is a key mediator of acute and chronic systemic inflammatory responses, not only inducing its own secretion but also stimulating the production of other inflammatory cytokines and chemokines.

[0045] Interleukin-6 (IL-6) is a pleiotropic cytokine belonging to the interleukin family. IL-6 is expressed by a single gene; its cDNA translation product is a 212-amino acid polypeptide chain, which can be cleaved to form a mature protein of 184 amino acids with a molecular weight of approximately 22-27 kDa. It is produced by fibroblasts, monocytes / macrophages, T lymphocytes, B lymphocytes, epithelial cells, keratinocytes, and various tumor cells. Interleukin-1, tumor necrosis factor-α, platelet-derived factor, and viral infections can all induce the production of IL-6 in normal cells. Interleukin-6 can stimulate the proliferation and differentiation of cells involved in the immune response and enhance their function. IL-6 not only has physiological activity on B cells but also on T cells, hematopoietic stem cells, hepatocytes, and brain cells. IL-6 is rapidly produced during acute inflammatory responses caused by internal or external injuries, stress responses, infections, and other conditions.

[0046] In this application, the biomarker triiodothyronine includes the free form of T3 (FT3) in the subject's serum, the biomarker Cortisol includes the free form of Cortisol in the subject's peripheral blood, the biomarker TNF-α includes the soluble mature TNF-α in the subject's peripheral blood, and the biomarker IL-6 includes the mature IL-6 in the subject's peripheral blood.

[0047] In a preferred embodiment of this application, the scope of biomarkers is further expanded to construct a more comprehensive and accurate diagnostic prediction model for depression. Endocrine indicators and / or cytokine indicators are incorporated, and the specific types of optional indicator components are increased. The diagnostic prediction model constructed using the data corresponding to the above indicator components can further improve the accuracy of diagnosis for the target population.

[0048] In this application, the inventors discovered that by using multidimensional serum biomarkers, it is possible to overcome the limitations of a single indicator and more comprehensively capture the heterogeneous pathological characteristics of depressive disorders, providing more effective feature support for constructing accurate and reliable diagnostic models for depressive disorders.

[0049] In a preferred embodiment, the machine learning method includes any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost; preferably, before constructing the diagnostic model using any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, support vector machine, AdaBoost, DT decision tree, and Xgboost, the construction method further includes: dividing the dataset into a training set and a test set, constructing a classification model using the training set, and verifying the model's performance using the test set.

[0050] In a preferred embodiment of this application, to ensure the accuracy and generalization ability of the constructed depression diagnosis and prediction model, we adopted a rigorous dataset partitioning process. Specifically, before machine learning training, the original dataset was meticulously divided into two subsets: a training set and a test set. The training set is used for model learning and construction, while the test set is reserved for objective validation of the model's performance, avoiding overfitting and ensuring the model's practical application value. Typically, the dataset partitioning ratio takes into account sample size, class balance, and model robustness. A common ratio is 70%-80% of the data as the training set and the remaining 20%-30% as the test set, ensuring that the model can both fully learn the data features and independently validate its predictive ability.

[0051] During the model building phase, one or more of the following machine learning algorithms, including but not limited to, can be used for training:

[0052] 1. Logistic Regression Model

[0053] Logistic regression is a classic linear probability model widely used in binary or multi-class classification prediction problems. It directly outputs the probability of a sample belonging to a certain class through maximum likelihood estimation, offering advantages such as strong parameter interpretability, high computational efficiency, and sensitivity to linear relationships among features. In the context of depression diagnosis, logistic regression can calculate the probability of an individual having depression based on features such as serum biomarker levels, cognitive assessment scores, and demographic variables. It can also achieve a hard classification of "depressed / non-depressed" by setting a threshold, while providing a quantitative contribution of each feature to the risk of developing the disorder.

[0054] 2. Random Forest Model

[0055] Random forest is a bagging-based ensemble learning method that improves predictive performance by constructing a large number of decision trees and voting or averaging their outputs. This algorithm can automatically capture complex nonlinear interactions and higher-order relationships between features, exhibits strong robustness to missing and outlier values, and can help discover key biomarkers through feature importance assessment. In the task of diagnosing depression, random forest can identify potential patterns among high-dimensional biomarkers and clinical variables that may interact, significantly improving the model's ability to discriminate depressive states and reducing the risk of overfitting.

[0056] 3. XGBoost Model (eXtreme Gradient Boosting Model)

[0057] XGBoost is a high-performance gradient boosting decision tree algorithm based on the Boosting framework. It iteratively trains weak learners and combines them in a weighted manner to gradually reduce the residual of the loss function. Its advantages include support for custom loss functions, built-in regularization to prevent overfitting, and efficient handling of sparse data. In depression diagnosis applications, XGBoost can deeply explore the nonlinear and non-monotonic relationship between biomarkers and depression, while outputting feature importance rankings to help researchers discover new potential pathogenic factors and achieve high accuracy and excellent generalization performance in depression risk prediction.

[0058] During the model training phase, the model will be trained using data from the training set and corresponding label information, employing one or more of the aforementioned algorithms to identify and learn the intrinsic association between biomarkers and depression. The goal of model training is to find an optimal set of parameters that maximizes the model's prediction accuracy on the training set while also considering its generalization ability and avoiding overfitting.

[0059] After model training is complete, the model's performance is objectively validated using a reserved test set. The test set includes independent samples that did not participate in model training. By comparing the model's predictions with actual diagnoses of depression, we can evaluate key performance indicators such as accuracy, sensitivity, and specificity, ensuring that the model's performance on unknown data meets expectations and providing a solid basis for future applications in real clinical settings.

[0060] Through the description of the preferred embodiments above, this application not only elaborates on the specific steps of dataset partitioning and model construction, but also introduces the application of several key machine learning algorithms in the diagnosis of depression. The selection of these algorithms is based not only on their general effectiveness in classification problems, but also on their unique ability to handle complex and diverse biomarker datasets. Through a detailed explanation of the training and validation process, this application clearly demonstrates how, through scientific methods, it constructs and optimizes a diagnostic model capable of accurately identifying depressive states in non-perinatal adult women, providing clinicians with a rapid and accurate auxiliary diagnostic tool, and is expected to significantly improve the diagnostic accuracy and treatment efficiency of depression.

[0061] In a preferred embodiment, the non-perinatal woman is a non-perinatal woman aged 19-65; preferably, the biological sample is one of serum, plasma or whole blood.

[0062] In this application, "non-perinatal women aged 19-65" refers to all female individuals aged 19 to 65 (inclusive) who are not currently in a perinatal state.

[0063] In a second typical embodiment of this application, a method for detecting depression is provided. The method is used to diagnose whether a subject has depression or to detect the risk of a subject having depression. The subject is a non-perinatal woman. The method includes: i) collecting biological samples from the subject and detecting the content of biomarkers in the method for constructing the diagnostic model of depression in the biological samples to obtain characteristic data of the biomarkers in the subject's biological samples; ii) generating a diagnosis result of depression or a prediction result of depression risk based on the characteristic data of the biomarkers in the subject's biological samples. Optionally, the characteristic data of the biomarkers in the subject's biological samples are selected from the content of the biomarkers or the data after standardization of the content of the biomarkers.

[0064] Optionally, ii) above includes: ii-1) comparing the levels of the subject's biomarkers with a first predetermined threshold, and generating a diagnosis of depression or a prediction of depression risk based on the comparison result; or ii-2) inputting the standardized data of the subject's biomarker levels into a pre-established depression diagnostic model, and comparing the calculated result with a second predetermined threshold, and generating a diagnosis of depression or a prediction of depression risk based on the comparison result; preferably, the first predetermined threshold is a threshold obtained based on an ROC curve not directly generated by a machine learning model, or a threshold obtained based on an ROC curve generated by a machine learning model; preferably, the second predetermined threshold is a prediction probability threshold or a model dependent variable threshold of a pre-established diagnostic model. Preferably, the diagnostic model is a diagnostic model obtained using the above-mentioned method for constructing a diagnostic model for depression.

[0065] When using a single diagnostic biomarker of 25-OH VD for diagnosis, the first predetermined threshold is selected from 16.8-18.8 ng / mL; for example, when using a single diagnostic biomarker of 25-OH VD for diagnosis, the threshold obtained based on the ROC curve generated directly without a machine learning model is 18.8 ng / mL; the threshold obtained based on the ROC curve generated based on the logistic regression equation is 16.8 ng / mL.

[0066] Optionally, the diagnostic model is constructed using machine learning based on known diagnoses of depression in non-perinatal women and the levels of biomarkers in biological samples from these women. Preferably, the machine learning method includes any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost. The results can be used to guide treatment plans.

[0067] "Predetermined threshold" refers to the parameter used to compare the marker or combination of markers in the subject's sample with the predetermined threshold when diagnosing the risk of disease, and output the subject's risk of disease or disease course based on the comparison result.

[0068] Specifically, the predetermined thresholds include, but are not limited to, positive judgment values, predicted probability thresholds, dependent variable thresholds, and other critical values ​​that can be used to divide results and clarify judgment boundaries. The setting of these thresholds should be based on the performance requirements of the target detection scenario (such as sensitivity, specificity, accuracy, etc.) and determined through clinical data validation, statistical model analysis (such as ROC curve analysis combined with Youden index calculation, etc.) or industry standard calibration, so as to ensure their judgment effectiveness and reliability in the detection method or diagnostic model.

[0069] Specifically, the characteristic data of biomarkers in the subject sample may include the biomarker content, or the result of standardizing the biomarker content.

[0070] Specifically, the predetermined threshold includes a first predetermined threshold and a second predetermined threshold.

[0071] In some specific implementations, the first predetermined threshold or the second preset threshold is obtained based on the principle of maximizing the Yoden index.

[0072] Specifically, in this paper, the first predetermined threshold is the positive cutoff value. The positive cutoff value refers to the critical single concentration value of the biomarker determined based on clinical sensitivity and specificity requirements when using ROC curves or models for data fitting. In one specific implementation, the biomarker concentration is directly compared with the positive cutoff value, and the result of the comparison is used to output the diagnosis or risk prediction of depression for the subject. For example, when the biomarker is 25-OH VD, the detection concentration value of 25-OH VD is directly compared with the positive cutoff value, and the result of the comparison is used to output the diagnosis or risk prediction of depression for the subject. Another example is when the biomarker is selected from the following combinations: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH VD + IL-6; 6) 25-OH VD + IL-6 + TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + cortisol + IL-6; 12) 25-OH VD + cortisol + IL-6 + TNFα; 13) 25-OH VD + triiodothyronine + cortisol + TNFα; 14) 25-OH VD + triiodothyronine + cortisol + TNFα. VD+triiodothyronine+cortisol+IL-6;15)25-OH VD+triiodothyronine+cortisol+IL-6+TNFα, according to each marker in the combination, a corresponding first predetermined threshold is set, and after comparison, the diagnosis or risk prediction result of depression of the subject is output according to the comparison result.

[0073] Specifically, the second predetermined threshold in this paper can be a prediction probability threshold, which refers to the prediction probability threshold (P-value threshold) of the model determined based on clinical sensitivity and specificity requirements when the diagnostic model is a logistic regression model, a random forest model, or other models. In one specific implementation, the biomarker feature data is substituted into the preset model to calculate the P-value of the corresponding biomarker feature data, and the P-value is compared with the model's prediction probability threshold. Based on the comparison result, the subject's depression diagnosis or risk prediction result is output.

[0074] The second predetermined threshold mentioned in this article can also be a dependent variable threshold, which refers to the preset dependent variable threshold of the model determined based on clinical sensitivity and specificity requirements when the diagnostic model is a functional formula. In one specific implementation, the biomarker feature data is substituted into the functional formula to calculate the dependent variable, and the dependent variable is compared with the dependent variable threshold of the functional formula. Based on the comparison result, the diagnosis or risk prediction result of depression for the subject is output.

[0075] The aforementioned method for detecting depression aims to achieve accurate diagnosis and risk assessment of depression by detecting the levels of specific biomarkers in non-perinatal women aged 19-65, optionally combined with advanced machine learning techniques. The core advantages of this method lie in its objectivity, speed, and high accuracy. It not only significantly improves the efficiency of clinical diagnosis but also provides patients with personalized risk assessments, offering a scientific basis for treatment planning.

[0076] First, this method involves collecting biological samples from the subject, such as serum, plasma, or whole blood samples. The sample collection process must follow standardized procedures to ensure the consistency and reliability of the results. This step is crucial for obtaining accurate biomarker data.

[0077] Subsequently, highly precise detection technology was used to quantitatively analyze the indicator components in the biological samples. In this embodiment, a series of biomarkers closely associated with depression in a specific target population were discovered, including but not limited to any combination of the following: 1) 25-OH vitamin D + triiodothyronine; 2) 25-OH vitamin D + cortisol; 3) 25-OH vitamin D + triiodothyronine + cortisol; 4) 25-OH vitamin D + TNFα; 5) 25-OH vitamin D + IL-6; 6) 25-OH vitamin D + IL-6 + TNFα; 7) 25-OH vitamin D + triiodothyronine + TNFα; 8) 25-OH vitamin D + triiodothyronine + IL-6; 9) 25-OH vitamin D + triiodothyronine + IL-6 + TNFα; 10) 25-OH vitamin D + cortisol + TNFα; 11) 25-OH vitamin D + cortisol + IL-6; 12) 25-OH vitamin D + cortisol + IL-6 + TNFα; 13) 25-OH vitamin D + triiodothyronine + IL-6 + TNFα; VD+triiodothyronine+cortisol+TNFα; 14)25-OHVD+triiodothyronine+cortisol+IL-6; 15)25-OHVD+triiodothyronine+cortisol+IL-6+TNFα.

[0078] After collecting and testing biological samples, the collected biomarker data can optionally be analyzed using the machine learning diagnostic prediction model constructed in the first embodiment of this application. This model is trained based on data from a certain number of sample populations, including healthy control groups and patients with depression. The strict division of the training and test sets ensures the model's generalization ability and prediction accuracy.

[0079] In practice, the content data of indicator components in the subject's biological sample are input into the model. Based on the mapping relationships learned during training, the model calculates the probability or risk level of the subject's depression. This process is not only fast but also avoids subjective bias, providing an objective basis for clinical decision-making.

[0080] The aforementioned method for detecting depression overcomes the limitations of traditional diagnostic methods by utilizing multi-dimensional biomarker detection and machine learning model prediction. It not only provides more objective and accurate diagnostic results than scale assessments, but also evaluates the risk of developing depression based on the subject's biomarker characteristics, offering a new approach for early screening and prevention. Furthermore, this method is specifically designed for non-perinatal adult women (aged 19-65), taking into account the unique endocrine and metabolic characteristics of this group, enabling it to more accurately reflect the biological characteristics of depression and opening new prospects for personalized medicine and precision diagnosis.

[0081] In a third typical embodiment of this application, a detection reagent is provided for use in the preparation of a product for diagnosing depression in non-perinatal women. The detection reagent is capable of detecting the content of biomarkers in biological samples from non-perinatal women, including 25-OH VD.

[0082] In a preferred embodiment, the biomarker further includes one or more of triiodothyronine, cortisol, TNFα, or IL-6; optionally, the biomarker includes any combination of the following: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH VD + IL-6; 6) 25-OH VD + IL-6 + TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + cortisol + IL-6; 12) 25-OH VD + cortisol + IL-6 + TNFα; 13) 25-OH VD + triiodothyronine + IL-6 + TNFα; VD+triiodothyronine+cortisol+TNFα; 14)25-OH VD+triiodothyronine+cortisol+IL-6; 15)25-OH VD+triiodothyronine+cortisol+IL-6+TNFα.

[0083] In a preferred embodiment, the method for detecting the content of the biomarker includes one or more of enzyme-linked immunosorbent assay (ELISA), chemiluminescent immunoassay, serum biochemical assay, high-performance liquid chromatography (HPLC), or mass spectrometry (MS); preferably, the product includes a kit or electronic predictive device; preferably, the detection reagent includes any one or more of the following: an antibody capable of binding to the biomarker, a tracer for labeling the antibody, a standard for the biomarker, a solid-phase carrier for antibody-antigen binding reaction, or a buffer solution. Those skilled in the art can flexibly configure relevant detection reagents according to the actual detection method and existing technical principles.

[0084] Those skilled in the art can use in vitro non-perinatal female depression diagnostic markers (markers with a certain concentration or proportion) or peripheral blood containing the markers as auxiliary diagnostic materials for non-perinatal female depression or as calibrators or quality control materials for diagnostic products.

[0085] The aforementioned applications include, but are not limited to, developing detection reagents or devices for detecting the levels of the aforementioned biomarkers, enabling precise detection of biomarkers, and facilitating the preparation of products for diagnosing depression in non-perinatal women. This innovation aims to fill the current gap in biomarker detection products for depression, providing a more objective and personalized diagnostic and risk assessment tool, with the goal of improving the clinical efficiency of diagnosis and treatment of depressive disorders.

[0086] Preferred detection methods include highly sensitive immunoassay techniques, mass spectrometry, enzyme-linked immunosorbent assay (ELISA), or detection techniques based on fluorescence, chemiluminescence, etc., to ensure accurate quantification of biomarkers.

[0087] The product can be designed as a kit or an electronic predictive device. Kits typically include all the reagents and tools needed to detect the aforementioned biomarkers, along with detailed usage instructions, suitable for routine testing in laboratories or clinics. Electronic predictive devices go a step further, integrating biomarker detection with data processing and predictive analysis. These may include handheld detectors, integrated analyzers, or cloud-connected intelligent diagnostic platforms, capable of analyzing test data in real time and outputting a probability or risk assessment of depression.

[0088] In a preferred embodiment, obtaining a diagnosis of depression in non-perinatal women based on the content of biomarkers includes: inputting the content of biomarkers into a diagnostic model to obtain a target value; comparing the target value with a preset diagnostic threshold to obtain a diagnosis of depression in non-perinatal women; wherein, the diagnostic model is constructed based on known depression diagnoses in the non-perinatal women population and the biomarker content in biological samples of the non-perinatal women population using any one of the following methods: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, and Xgboost.

[0089] In a fourth typical embodiment of this application, an electronic device is provided for diagnosing whether a non-perinatal adult woman suffers from depression, or for detecting the risk of depression in a non-perinatal adult woman. The electronic device includes a data collection module and a diagnostic module. The data collection module collects the levels of biomarkers in a subject's biological sample and inputs the data collected by the data collection module into the diagnostic module. The biomarkers include 25-OH. VD; Diagnostic module, configured to output diagnostic or auxiliary diagnostic results or risk prediction results for the subject based on the biomarker content in the subject's biological sample; Optionally, the diagnostic module contains a diagnostic model, which calculates and outputs a judgment result using the data collected by the data collection module and the diagnostic model; The diagnostic model is constructed based on the known diagnosis of depression in non-perinatal women and the biomarker content in the biological sample of non-perinatal women using any one of the following methods: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, and Xgboost; Optionally, the biomarker also includes one or more of triiodothyronine, cortisol, TNFα, or IL-6; Optionally, the biomarker includes any combination of the following: 1) 25-OH VD + triiodothyronine; 2) 25-OH VD + cortisol; 3) 25-OH VD + triiodothyronine + cortisol; 4) 25-OH VD + TNFα; 5) 25-OH 6) 25-OH VD+IL-6+TNFα; 7) 25-OH VD+Triiodothyronine+TNFα; 8) 25-OH VD+Triiodothyronine+IL-6; 9) 25-OH VD+Triiodothyronine+IL-6+TNFα; 10) 25-OH VD+Cortisol+TNFα; 11) 25-OH VD+Cortisol+IL-6; 12) 25-OH VD+Cortisol+IL-6+TNFα; 13) 25-OH VD+Triiodothyronine+Cortisol+TNFα; 14) 25-OH VD+Triiodothyronine+Cortisol+IL-6; 15) 25-OH VD+Triiodothyronine+Cortisol+IL-6+TNFα.

[0090] Optionally, the diagnostic module uses the content of indicator components and the diagnostic model to calculate and output the judgment result, including: inputting the content of biomarkers into the diagnostic model to obtain the target value; comparing the target value with a predetermined threshold to obtain the diagnosis result or risk prediction result of depression in non-perinatal women.

[0091] Optionally, the diagnostic module uses the content of the biomarkers and the diagnostic model to calculate and output the judgment result, including: inputting the content of the biomarkers into the diagnostic model after standardization to obtain a target value; comparing the target value with a preset diagnostic threshold to obtain the diagnosis result or risk prediction result of depression in the non-perinatal women.

[0092] Preferably, the diagnostic model described above is a diagnostic model obtained using the method for constructing the diagnostic model of depression described above.

[0093] This application proposes an innovative electronic device specifically for the diagnosis of depression in non-perinatal adult women.

[0094] The data collection module is the front end of the device, responsible for collecting and analyzing the subject's biological samples to detect the content of specific biomarkers, or for collecting data from biological samples obtained from other detection devices.

[0095] The diagnostic module is the core computing unit of the device, responsible for outputting a diagnosis or risk prediction result for the subject based on the biomarker levels in the subject's biological samples. Optionally, the diagnostic module incorporates a machine learning-based diagnostic model. This model is built upon a large amount of clinical sample data and trained using algorithms such as logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost, enabling accurate identification of the risk of depression in non-perinatal adult women. When the data collection module transmits the detected biomarker levels to the diagnostic module, the module immediately initiates the diagnostic procedure, using the built-in model to perform in-depth analysis of the data, calculate the probability or risk level of the subject's depression, and ultimately output the judgment result. This process is fast and efficient, significantly shortening the diagnostic cycle and improving the timeliness of clinical decision-making.

[0096] In a fifth typical embodiment of this application, a method for detecting depression is provided. The steps of the method are performed by a computer. The detection method is used to diagnose whether a subject has depression or to detect the risk of depression in a subject. The subject is a non-perinatal woman. The detection method includes: generating a diagnosis result of depression or a prediction result of depression risk for the subject based on the characteristic data of biomarkers in the subject's biological sample; optionally, the characteristic data of biomarkers in the subject's biological sample is selected from the content of the biomarkers or the data of the biomarkers after standardization; optionally, the detection method includes: comparing the content of the subject's biomarkers with a first predetermined threshold, and generating a diagnosis result of depression or a prediction result of depression risk for the subject based on the comparison result; or inputting the standardized data of the subject's biomarkers into a pre-established depression diagnosis model, and comparing the calculated result with a first predetermined threshold. A second predetermined threshold is compared, and based on the comparison result, a diagnosis result of depression or a prediction result of depression risk for the aforementioned subjects is generated; preferably, the first predetermined threshold is a threshold obtained based on an ROC curve generated directly without a machine learning model, or a threshold obtained based on an ROC curve generated by a machine learning model; preferably, the second predetermined threshold is a predicted probability threshold or a dependent variable threshold of a pre-established diagnostic model; the diagnostic model is constructed by machine learning based on known depression diagnoses of non-perinatal women and the content of the aforementioned biomarkers in biological samples of the non-perinatal women; preferably, the machine learning includes any one of logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost; preferably, the diagnostic model is a diagnostic model obtained using the aforementioned method for constructing a diagnostic model of depression.

[0097] In a sixth typical embodiment of this application, the present invention also provides a treatment method for depression in non-perinatal women, comprising the following steps: diagnosing or predicting the risk of depression in subjects using the non-perinatal women depression detection method described in any of the foregoing embodiments or implementations, and obtaining the diagnosis or risk prediction results; and providing corresponding treatment to confirmed or high-risk patients based on the depression diagnosis or risk prediction results. In optional embodiments, the above treatment methods include any one or more of cognitive behavioral therapy (CBT), interpersonal therapy (IPT), psychodynamic therapy, mindfulness therapy (such as mindfulness-based stress reduction (MBSR) and mindfulness-based cognitive therapy (MBCT), drug therapy, and physical therapy. In optional embodiments, the above drug therapy includes any one or more of selective serotonin reuptake inhibitors (SSRIs), serotonin and norepinephrine reuptake inhibitors (SNRIs), norepinephrine and specific serotonergic antidepressants (NaSSAs), and other classes of drugs. In optional implementations, SSRIs include any one or more of fluoxetine (Prozac), paroxetine, sertraline (Zoloft), fluvoxamine, citalopram, and escitalopram. SNRIs include venlafaxine and duloxetine. NaSSAs include mirtazapine. Other classes of drugs include bupropion and agomelatine.

[0098] In a seventh typical embodiment of this application, a kit is provided, comprising a detection reagent and instructions; the detection reagent includes any one of the following reagents: A, a detection reagent for detecting 25-OH VD biomarker levels; B, a detection reagent for detecting 25-OH VD biomarker levels; and a detection reagent for detecting at least one biomarker selected from triiodothyronine, Cortisol, IL6, or TNFα; the instructions indicate how to use the biomarker levels corresponding to the detection reagents for the diagnosis or risk prediction of depression in non-perinatal women.

[0099] In an eighth typical embodiment of this application, a computer-readable storage medium is provided, the storage medium including a stored program, wherein, when the program is running, the device where the storage medium is located executes the method for constructing the diagnostic model of depression or the method for detecting depression.

[0100] In a ninth typical embodiment of this application, a processor is provided for running a program, wherein the program executes the method for constructing the diagnostic model of depression or the method for detecting depression.

[0101] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the present invention.

[0102] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus hardware devices such as detection devices. Based on this understanding, the data processing part of the technical solution of this application can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of the embodiments of this application.

[0103] This application can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0104] Obviously, those skilled in the art should understand that some modules or steps of this application described above can be implemented in general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps into a single integrated circuit module. Thus, this application is not limited to any particular hardware and software combination.

[0105] The beneficial effects of this application will be explained in more detail below with reference to specific embodiments.

[0106] Example

[0107] 1. Detection methods for markers

[0108] The assays used in this example included the detection of 25-OH vitamin D (catalog number 130261004M), FT3 (free triiodothyronine, catalog number 1302533005M), Cortisol (cortisol, catalog number 130270002M), TNFα (tumor necrosis factor-α, catalog number 130216006M), IL-6 (interleukin-6, catalog number 130216504M), ADPN (adiponectin, catalog number 130505009S), LAC (lactic acid, catalog number 130505003SSC801), TC (total cholesterol, catalog number 130551003SSC801), TG (triglycerides, catalog number 130551004SSC801), and T. The kits for SH (thyroid-stimulating hormone, catalog number: 130253001M), E2 (estradiol, catalog number: 130252007M), TEST (testosterone, catalog number: 1302523001M), ALD (aldosterone, catalog number: 130256007M), CRP (C-reactive protein, catalog number: 130266002M), SAA (serum amyloid A, catalog number: 130216005M), NSE (neuron-specific enolase, catalog number: 130251030M), and S100 (S100 protein, catalog number: 130251017M) were sourced from Shenzhen New Industries Biomedical Engineering Co., Ltd., and the tests were performed according to the kit instructions.

[0109] 2. Verification of the role of biomarkers in the auxiliary diagnosis of depressive disorders

[0110] 2.1 Discovery of biomarkers for the auxiliary diagnosis of depressive disorders

[0111] (1) Research sample information:

[0112] Sample type: A sample refers to a human peripheral blood sample, including plasma or serum.

[0113] Sampling method: 5 mL of fasting venous blood was collected from patients with depressive disorders and healthy individuals using the well-known negative pressure blood collection technique. The serum was separated by centrifugation at 3000 rpm for 10 min and stored at -80℃ for testing.

[0114] Sample source and population information: Samples were collected continuously from Hospital A (samples collected continuously within a certain period of time).

[0115] Peripheral blood samples, excluding perinatal individuals, included serum samples from 45 patients with depressive disorders and 32 healthy individuals undergoing physical examinations. They were divided into a depression group and a healthy group. The median age of the healthy group was 24 years, and the median age of the depression group was 19 years. There were no statistically significant differences in age and sex between the two groups (P = 0.333 and 0.632, respectively).

[0116] (2) Experimental methods

[0117] The concentrations of potential biomarkers in samples from the depression group and the healthy group were detected using the aforementioned test kit and related detection methods. The tests were performed according to the instructions provided with the kit, and the differences in potential biomarker levels between the depression group and the healthy group were compared.

[0118] The results of the differential analysis of potential biomarkers between the depression group and the healthy group are shown in Table 1.

[0119] Table 1

[0120]

[0121] As shown in Table 1, among the 17 potential biomarkers, the levels of metabolic marker 25-OH VD, endocrine markers FT3 and cortisol, and inflammatory and cytokine markers TNFα, IL-6, and ADPN showed significant differences between the healthy group and the depressed group, indicating that these biomarkers can be used as predictive model variables.

[0122] 2.2 A preliminary study on the role of 25-OH vitamin D in the auxiliary diagnosis of depressive disorders in a small sample size of adult women.

[0123] (1) Research sample information:

[0124] Sample type: A sample refers to a human peripheral blood sample, including plasma or serum.

[0125] Sampling method: 5 mL of fasting venous blood was collected from patients with depressive disorders and healthy individuals using the well-known negative pressure blood collection technique. The serum was separated by centrifugation at 3000 rpm for 10 min and stored at -80℃ for testing.

[0126] Sample source and population information: Peripheral blood samples were collected continuously from adult women aged 19-65 years from Hospital A (samples collected continuously within a certain period of time), excluding the perinatal population. Serum samples were collected from 41 patients with depressive disorders and 33 healthy individuals undergoing physical examinations. They were divided into a depressive disorder group and a healthy control group. The median age of the two groups was 37 and 30 years, respectively, with no significant difference. P =0.194). It should be noted that although the above samples and the samples in 2.1 both come from the same hospital, there is no overlap or duplication between them.

[0127] Experimental methods:

[0128] The concentrations of 25-OH VD in samples from the depressive disorder group and the healthy control group were detected using the 25-OH VD detection kit and method provided in the examples. The detection was performed according to the instructions provided with the aforementioned kit. The differences in 25-OH VD levels between the depressive disorder group and the healthy control group were compared, and ROC curve analysis was used to evaluate the performance of 25-OH VD in distinguishing between patients with depressive disorders and healthy individuals.

[0129] (2) Analysis of differences in 25-OH vitamin D between the depression group and the healthy group

[0130] The differences in 25-OH vitamin D levels between the depressive disorder group and the healthy group were analyzed, and the results are as follows: Figure 1 As shown, the 25-OH vitamin D level in the adult female depression group was significantly lower than that in the healthy control group. P <0.0001).

[0131] (3) ROC curve analysis results of 25-OH VD in diagnosing depressive disorders

[0132] Using 25-OH vitamin D to differentiate between the healthy control group and the depressive disorder group, the ROC analysis results are as follows: Figure 2 As shown, the ROC curve AUC of 25-OH VD distinguishing between the healthy control group and the depressive disorder group was 0.751. The optimal threshold was 18.8 ng / mL, the sensitivity was 0.84, and the specificity was 0.65, demonstrating that 25-OH VD has high diagnostic performance as an auxiliary diagnostic biomarker for depressive disorders in adult women.

[0133] 3. A study on the role of a logistic regression model constructed from 25-OH vitamin D and its biomarkers in the diagnosis of depressive disorders in adult women.

[0134] 3.1 Provide the research population, sample source, experimental methods, and data analysis methods.

[0135] (1) Research sample information:

[0136] Sample distribution study at different disease stages: Information on patients with depression admitted to multiple hospitals was collected, and the proportions of patients at different disease stages were statistically analyzed, as shown in Table 2.

[0137] Table 2

[0138]

[0139] Based on the epidemiological findings regarding the gender and age distribution of the population, as well as the proportions of different disease courses, serum samples were continuously collected from adult female patients with depressive disorders and healthy individuals from Hospital C, excluding the perinatal population. The samples included 90 patients with depressive disorders and 51 healthy individuals undergoing physical examinations. They were divided into a depression group and a healthy group. The age and disease distribution of the two groups are shown in Tables 3 and 4 below.

[0140] Table 3

[0141]

[0142] Both the healthy group and the depressed group covered people of different age groups. Statistically, there was no significant difference in the age distribution between the healthy group and the depressed group according to the chi-square test.

[0143] Table 4

[0144]

[0145] The patient group covered patients with depression of different severities, including mild to moderate depressive episodes and severe depressive episodes. Among them, patients with severe depressive episodes included those without psychotic symptoms and those with psychotic symptoms. The proportion of consecutive sampling did not show significant differences from the statistical results of multiple hospitals, and the sample collection was terminated.

[0146] Sample type: A sample refers to a human peripheral blood sample, including plasma or serum.

[0147] Sampling method: 5 mL of fasting venous blood was collected from patients with depressive disorders and healthy individuals using the well-known negative pressure blood collection technique. The serum was separated by centrifugation at 3000 rpm for 10 min and stored at -80℃ for testing.

[0148] (2) Data analysis methods:

[0149] Statistical analysis was performed using R-4.5.1 software. Normally distributed continuous data were analyzed using... ± This indicates that the comparison between the two groups uses... t Tests: Non-normally distributed continuous data are expressed as M (interquartiles), and the Mann-Whitney U test is used for comparisons between two groups; chi-square test is used for count / categorical data (e.g., age distribution, disease duration distribution). P <0.05 indicates statistical significance; the area under the receiver operating characteristic (ROC) curve (AUC) was used to assess the discriminative power of the predictive biomarkers, and the sensitivity and specificity at the optimal cutoff value were calculated.

[0150] 3.2 Research on the role of 25-OH vitamin D in the auxiliary diagnosis of depressive disorders

[0151] (1) Experimental methods

[0152] The concentrations of 25-OH VD in samples from adult women with depressive disorders and healthy controls were detected using the 25-OH VD detection kit and method provided in the examples. The detection was performed according to the instructions provided with the aforementioned kit. The differences in 25-OH VD levels between the depressive disorder group and the healthy control group were compared, and the results are shown in Table 5.

[0153] (2) Analysis of differences in 25-OH VD between the depressive disorder group and the healthy group

[0154] Table 5

[0155]

[0156] 3.3 A study on the role of biomarker combinations in constructing logistic regression models for the diagnosis of depressive disorders in adult women.

[0157] (1) Model building method

[0158] A logistic regression diagnostic model was constructed using a combination of various biomarkers (25-OH VD, FT3, cortisol, TNFα, and IL-6). Healthy controls and depressive disorder groups among adult women were randomly assigned to training and testing sets in a 3:1 ratio. The training set was used to build the model, and the testing set was used to evaluate its performance. Before constructing the logistic regression model, the biomarkers were standardized using Z-scores to eliminate dimensions.

[0159] Note: Data standardization methods also include min-ma normalization, robust scaling, etc., and are not limited to Z-score standardization.

[0160] The formula for calculating the prediction probability of a logistic regression model is: P = 1 / (1 + e^(-1 / 2)) -Y (e is the natural constant, Y is the linear formula, and P is the predicted probability). Based on the model's predicted probability, the AUC is calculated using ROC curve analysis, and the threshold corresponding to the maximum Youden index is selected as the predicted probability threshold. The model's sensitivity and specificity are calculated to evaluate the model's ability to distinguish between healthy individuals and depressed patients. (Note: Logistic regression is one example, but it is not limited to logistic regression; it also includes well-known model building methods such as random forest and XGboost.)

[0161] (2) Results of training and testing sets

[0162] The results for the training and testing sets are shown in Tables 6 and 7.

[0163] Table 6. Logistic Regression Models and Model Discriminant Values ​​for Each Biomarker Combination in the Training Set

[0164]

[0165] Table 7

[0166]

[0167] Based on Tables 5, 6, and 7, it can be seen that there are significant differences in 25-OH VD between depressed patients and healthy individuals in the sample provided in 3.1, which is basically consistent with the conclusions of the sample provided in 2.1. Furthermore, the AUC of 25-OH VD on the test set was verified to be 0.72, indicating that this indicator has good diagnostic value in adult women.

[0168] For each combined assay protocol involving 25-OH VD with one or more of the provided endocrine or cytokine markers, the AUC of both the training and validation sets was greater than 0.7, indicating that the combined assay protocols provided in this study can be used for the diagnosis of depressive disorders in adult women. Since the training and validation sets were obtained through random sampling and covered adult female patients of different ages and disease stages, this also demonstrates the good generalization ability of the provided assay protocols in adult women, enabling them to provide diagnostic evidence for adult women with depressive disorders of different ages and disease stages.

[0169] It should be noted that although some of the above data show sensitivity or specificity below 0.6, this is only a cutoff value based on the principle of maximizing the Youden index. Clinically, the cutoff value and the corresponding balance between sensitivity and specificity can be reselected within the scheme provided by this invention according to the needs.

[0170] The data for the test set includes:

[0171] The combined detection protocols for 25-OH vitamin D and FT3 (AUC > 0.8) and 25-OH vitamin D and cortisol (AUC > 0.75) were superior to the combined detection protocols for 25-OH vitamin D and cytokines (AUC: 0.71-0.72). This indicates that changes in 25-OH vitamin D-related metabolic factors and changes in FT3 and cortisol-related hormone levels are important factors in the diagnosis of depression in adult women.

[0172] Furthermore, the combined detection regimens of 25-OH VD and FT3, and 25-OH VD and cortisol, showed an improvement of more than 0.05 compared to their respective individual detection regimens, which is higher than the improvement of other combined detection indicators. This indicates that the two indicators, 25-OH VD and FT3, and 25-OH VD and cortisol, have a good complementary effect in diagnosing depressive disorders in adult women.

[0173] The three-item combined detection scheme for 25-OH VD, FT3, and cortisol (AUC=0.87) is superior to the aforementioned schemes that detect only two items each, with an improvement of ≥0.05, indicating that the three items of 25-OH VD, FT3, and cortisol also have complementary effects.

[0174] Based on the above results, we speculate that this may be related to the unique pathogenic mechanisms of depression in adult women.

[0175] 25-OH vitamin D serves as a core biomarker for immune homeostasis and neural repair; FT3 is a key indicator of systemic and central energy metabolism and homeostasis; and Cortisol is a classic indicator of stress-induced neuroendocrine responses. These three may each dominate an independent pathophysiological axis in adult women, with strong bidirectional interactions between them. Our data suggest that dysregulation of any two axes can significantly improve diagnostic efficacy, while assessing all three axes simultaneously maximizes the capture of the most complete biological heterogeneity in adult women with depression, thus yielding the best diagnostic results.

[0176] Therefore, we propose a model in which the “synergistic dysregulation” of the three major axes of immunity (VD), metabolism (FT3), and stress (Cortisol) is the core biological feature of depression in adult women. This provides a new perspective for understanding the disease mechanism and developing targeted therapies in this population.

[0177] Based on the above results, we speculate that this may be related to the heterogeneity of the adult female population.

[0178] Vitamin D (VD) plays a protective role by regulating glial cell activity and neurotransmitter metabolism, thereby inhibiting neuroinflammation. Triiodothyronine (FT3) affects nerve metabolism and the adaptive remodeling of synapses through its energy regulation of nerve cells. Cortisol reflects the state of the body's stress regulation axis (HPA axis) and is closely related to mood regulation. Adult women are affected by estrogen fluctuations and specific physiological stages (such as menstruation, pregnancy, and menopause), which alter the activity of VD receptors, the efficiency of thyroid hormone conversion in the periphery, and the sensitivity of the HPA axis. This makes the mechanisms of action of these indicators vary from person to person. Therefore, the combined detection of 25-hydroxyvitamin D (25-OH VD) and FT3 / cortisol can capture more pathological information about depression in adult women, avoiding missed or misdiagnosed cases caused by a single indicator. Simultaneous assessment of these three indicators can more comprehensively reflect the individual physiological differences in adult women with depression, thus achieving the best diagnostic results.

[0179] Adding cytokine-related indicators to 25-OH VD and FT3 resulted in a detection value of no less than 0.8. Adding cytokine-related indicators to 25-OH VD and cortisol also resulted in a detection value of no less than 0.75, thus maintaining diagnostic value.

[0180] Based on the three-item combined detection scheme provided by this invention, additional detection indicators within the scope provided by this invention are added, and their AUCs are all greater than 0.8. Therefore, the four-item and more-than-four-item schemes provided by this invention also have good diagnostic value.

[0181] (3) Diagnostic performance verification based on support vector machine.

[0182] Based on the enrolled sample of the study "The role of logistic regression model constructed from 3, 25-OH VD and biomarker combinations in the diagnosis of depressive disorders in adult women", support vector machine models were constructed for single biomarkers or combinations of biomarkers 25-OH VD, FT3, Cortisol, IL6, and TNFα.

[0183] (a) Dataset partitioning and preprocessing: The healthy group and the depressed group were randomly divided into training and test sets in a 3:1 ratio. The training set was used for model training, and the test set was used to evaluate the model's performance. The YeoJohnson method was used to perform normalization transformation on the quantitative variables. The Z-score method was used to standardize the quantitative variables.

[0184] (b) The kernel function for the support vector machine model is chosen to be a radial basis function (RBF). The hyperparameters to be tuned include Cost and RBF_sigma (the range of influence of a single training sample in the feature space). The optimal hyperparameters are searched using a grid search method, and AUC is used as the tuning metric. The combination of hyperparameters with the highest average AUC after 10-fold cross-validation is selected.

[0185] (c) Model Evaluation: Based on the optimal hyperparameter combination, the model parameters are trained using the training set data, and the predicted probabilities of the training set are obtained. ROC curve analysis is used to determine the threshold for predicted classification. Based on the trained support vector machine model, predictions are made on the test set data to obtain the predicted probabilities and predicted classifications. AUC, sensitivity, and specificity are calculated to evaluate the model's ability to distinguish between the healthy group and the depressed group.

[0186] The diagnostic performance results of support vector machines are shown in Table 8 below.

[0187] Table 8

[0188]

[0189] Based on the test set data in Table 8, the AUC of the ROC curve distinguishing the healthy group from the depressed group by 25-OH VD was 0.72, demonstrating that 25-OH VD alone has high diagnostic performance as an auxiliary diagnostic marker for depression in non-perinatal adult women. The combined detection scheme of 25-OH VD with FT3 and cortisol showed a greater improvement in results compared to the individual detection schemes than the schemes with IL6 or TNFα alone, consistent with the logistic regression model conclusions. This indicates that 25-OH VD and FT3, as well as 25-OH VD and cortisol, have good complementary effects in diagnosing depressive disorders in non-perinatal adult women. The three-item combined detection scheme of 25-OH VD, FT3, and cortisol was superior to the aforementioned two-item combined detection schemes, with an improvement of ≥0.05, indicating that 25-OH VD, FT3, and cortisol also have complementary effects.

[0190] The combined detection of at least one of FT3, cortisol, IL6, and TNFα in 25-OH VD does not impair or even enhances the diagnostic value, consistent with the conclusions of the logistic regression model. This demonstrates that different model construction methods do not affect the diagnostic performance of the biomarkers and their combinations selected in this invention.

[0191] As can be seen from the above description, the embodiments of the present invention achieve the following technical effects: by utilizing 25-OH VD and optional other biomarkers, combined with a diagnostic model constructed through machine learning, the present invention can significantly improve the accuracy and efficiency of diagnosing depression in non-perinatal women. Compared with traditional clinical interviews and scale assessments, biomarker detection provides more objective and quantitative information, reduces errors in subjective judgment, and makes the diagnostic process more scientific and standardized.

[0192] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing a diagnostic model for depression, characterized in that, The diagnostic model is used to diagnose whether non-perinatal women suffer from depression. The construction method includes: a) Obtain the depression status and biological samples of the sample population, detect biomarkers in the biological samples, and establish a dataset of the biological samples. The dataset includes a mapping relationship between the prevalence of depression in the sample population and the levels of the biomarkers in the biological samples; b) Construct the diagnostic model using machine learning methods; The sample population included healthy women outside the perinatal period and patients with depression. Among them, biomarkers include 25-OH VD.

2. The construction method according to claim 1, characterized in that, The biomarkers also include one or more of triiodothyronine, cortisol, TNFα, or IL-6; Optionally, the biomarker includes any combination of the following: 1) 25-OH vitamin D + triiodothyronine; 2) 25-OH vitamin D + cortisol; 3) 25-OH vitamin D + triiodothyronine + cortisol; 4) 25-OH VD+TNFα; 5) 25-OH VD+IL-6; 6) 25-OH VD+IL-6+TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + Cortisol + IL-6; 12) 25-OH VD+cortisol+IL-6+TNFα; 13) 25-OH VD + triiodothyronine + cortisol + TNFα; 14) 25-OH VD + triiodothyronine + cortisol + IL-6; 15) 25-OH VD + triiodothyronine + cortisol + IL-6 + TNFα; Preferably, the triiodothyronine is free triiodothyronine.

3. The construction method according to claim 1 or 2, characterized in that, The machine learning methods include any one of the following: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost. Preferably, before constructing the diagnostic model using any one of the methods described above—logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, and Xgboost—the construction method further includes: dividing the dataset into a training set and a test set, constructing a classification model using the training set, and verifying the model's performance using the test set.

4. The construction method according to claim 1 or 2, characterized in that, The non-perinatal women referred to are women aged 19-65 who are not in the perinatal period; Preferably, the biological sample is one of serum, plasma, or whole blood.

5. A method for detecting depression, characterized in that, The detection method is used to diagnose whether a subject has depression or to detect the risk of a subject having depression. The subject is a non-perinatal woman. The detection method includes: i) Collect biological samples from the subject, detect the content of biomarkers in the method for constructing the diagnostic model of depression according to any one of claims 1-4 in the biological samples, and obtain characteristic data of biomarkers in the subject's biological samples; ii) Based on the characteristic data of biomarkers in the subject's biological samples, generate a diagnosis of depression or a risk prediction result for depressive disorder in the subject; Optionally, the characteristic data of the biomarkers in the subject's biological samples are selected from the content of the biomarkers or the standardized data of the content of the biomarkers; Optionally, ii) includes: ii-1) Compare the levels of the subject's biomarkers with a first predetermined threshold, and generate a diagnosis of depression or a prediction of depression risk for the subject based on the comparison result; or ii-2) The standardized data of the biomarker levels of the subject are input into a pre-established depression diagnosis model, and the calculated results are compared with a second predetermined threshold. Based on the comparison results, the depression diagnosis result or depression risk prediction result of the subject is generated. Preferably, the first predetermined threshold is a threshold obtained based on an ROC curve generated directly without a machine learning model, or a threshold obtained based on an ROC curve generated by a machine learning model; Preferably, the second predetermined threshold is a predicted probability threshold of a pre-established diagnostic model or a threshold of the model's dependent variable; Optionally, the diagnostic model is constructed using machine learning based on known diagnoses of depression in non-perinatal women and the levels of biomarkers in biological samples from the non-perinatal women population. Preferably, the machine learning includes any one of the following methods: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost. Preferably, the diagnostic model is a diagnostic model obtained using the method for constructing a diagnostic model for depression according to any one of claims 1-4.

6. The application of the test reagent in the preparation of products for diagnosing depression in non-perinatal women, characterized in that, The detection reagent enables the detection of biomarkers in biological samples from non-perinatal women, including 25-OH VD.

7. The application according to claim 6, characterized in that, The biomarkers also include one or more of triiodothyronine, cortisol, TNFα, or IL-6; Optionally, the biomarker includes any combination of the following: 1) 25-OH vitamin D + triiodothyronine; 2) 25-OH vitamin D + cortisol; 3) 25-OH vitamin D + triiodothyronine + cortisol; 4) 25-OH VD+TNFα; 5) 25-OH VD+IL-6; 6) 25-OH VD+IL-6+TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + Cortisol + IL-6; 12) 25-OH VD+cortisol+IL-6+TNFα; 13) 25-OH VD + triiodothyronine + cortisol + TNFα; 14) 25-OH VD + triiodothyronine + cortisol + IL-6; 15) 25-OH VD+ Triiodothyronine+ Cortisol+IL-6+TNFα.

8. The application according to claim 6 or 7, characterized in that, The detection methods for the content of the biomarkers include one or more of the following: enzyme-linked immunosorbent assay (ELISA), chemiluminescent immunoassay, serum biochemical assay, high performance liquid chromatography (HPLC), or mass spectrometry (MS). Preferably, the product includes a reagent kit or an electronic prediction device; Preferably, the detection reagent includes any one or more of the following: an antibody capable of binding to the biomarker, a tracer for labeling the antibody, a standard of the biomarker, a solid-phase carrier for antibody-antigen binding reaction, or a buffer solution.

9. An electronic device for diagnosing whether a non-perinatal adult woman suffers from depression, or for detecting the risk of depression in a non-perinatal adult woman, characterized in that, The electronic device includes a data collection module and a diagnostic module. The data collection module is used to collect the content of biomarkers in the biological samples of the subjects, and input the data collected by the data collection module into the diagnostic module; wherein the biomarkers include 25-OH VD; The diagnostic module is configured to output the subject's diagnostic results or risk prediction results based on the biomarker content in the subject's biological sample. Optionally, the diagnostic module includes a diagnostic model. The diagnostic module uses the data collected by the data collection module and the diagnostic model to calculate and output a judgment result. The diagnostic model is constructed based on the known diagnoses of depression in non-perinatal women and the levels of biomarkers in the biological samples of the non-perinatal women using any one of the following methods: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost. Optionally, the biomarker may further include one or more of triiodothyronine, cortisol, TNFα, or IL-6; Optionally, the biomarker includes any combination of the following: 1) 25-OH vitamin D + triiodothyronine; 2) 25-OH vitamin D + cortisol; 3) 25-OH vitamin D + triiodothyronine + cortisol; 4) 25-OH VD+TNFα; 5) 25-OH VD+IL-6; 6) 25-OH VD+IL-6+TNFα; 7) 25-OH VD + triiodothyronine + TNFα; 8) 25-OH VD + triiodothyronine + IL-6; 9) 25-OH VD + triiodothyronine + IL-6 + TNFα; 10) 25-OH VD + cortisol + TNFα; 11) 25-OH VD + Cortisol + IL-6; 12) 25-OH VD+cortisol+IL-6+TNFα; 13) 25-OH VD + triiodothyronine + cortisol + TNFα; 14) 25-OH VD + triiodothyronine + cortisol + IL-6; 15) 25-OH VD + triiodothyronine + cortisol + IL-6 + TNFα; Optionally, the diagnostic module uses the content of the biomarker and the diagnostic model to calculate and output a judgment result, including: The levels of the biomarkers are standardized and then input into the diagnostic model to obtain target values; The target value is compared with a predetermined threshold to obtain the diagnosis result or risk prediction result of depression in the non-perinatal women; Preferably, the diagnostic model is a diagnostic model obtained using the method for constructing a diagnostic model for depression according to any one of claims 1-4.

10. A method for detecting depression, characterized in that, The steps of the method are performed by a computer. The detection method is used to diagnose whether a subject has depression or to detect the risk of a subject having depression. The subject is a non-perinatal woman. The detection method includes: Based on the characteristic data of biomarkers in the subjects' biological samples, predictive results of the subjects' depression risk are generated. Optionally, the biomarker characteristic data of the subject's biological sample are selected from the biomarker content or the biomarker content after standardization. Optionally, the detection method includes: The levels of the subject's biomarkers are compared with a first predetermined threshold, and based on the comparison results, a diagnosis of depression or a prediction of depression risk for the subject is generated; or The standardized data of the subjects' biomarker levels are input into a pre-established depression diagnostic model, and the calculated results are compared with a second predetermined threshold. Based on the comparison results, a depression diagnosis result or a depression risk prediction result for the subjects is generated. Preferably, the first predetermined threshold is a threshold obtained based on an ROC curve generated directly without a machine learning model, or a threshold obtained based on an ROC curve generated by a machine learning model; Preferably, the second predetermined threshold is a predicted probability threshold of a pre-established diagnostic model or a threshold of the model's dependent variable; The diagnostic model was constructed using machine learning based on known diagnoses of depression in non-perinatal women and the levels of the biomarkers in biological samples from the non-perinatal women population. Preferably, the machine learning includes any one of the following methods: logistic regression, random forest, K-nearest neighbors, Gaussian Bayes, Naive Bayes, support vector machine, AdaBoost, DT decision tree, or Xgboost. Preferably, the diagnostic model is a diagnostic model obtained using the method for constructing a diagnostic model for depression according to any one of claims 1-4.

11. A reagent kit, characterized in that, The kit includes test reagents and instructions; The detection reagent includes any one of the following reagents: A. Detection reagents for detecting the level of 25-OH VD markers; B. A test reagent for detecting the level of 25-OH VD markers, and a test reagent for detecting the level of at least one of the markers triiodothyronine, Cortisol, IL6, or TNFα; The instructions specify how to use the biomarker levels corresponding to the test reagent for the diagnosis or risk prediction of depression in non-perinatal women.

12. A computer-readable storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, the device on which the storage medium is located is controlled to execute the method for constructing a diagnostic model for depression according to any one of claims 1-4, or the method for detecting depression according to claim 5, or the method for detecting depression according to claim 10.

13. A processor, characterized in that, The processor is used to run a program, wherein the program executes the method for constructing a diagnostic model for depression according to any one of claims 1-4, or the method for detecting depression according to claim 5, or the method for detecting depression according to claim 10.

Images