Plasma metabolic diagnostic marker and diagnostic system for depression in children and adolescents

Abstract

The present application relates to the technical field of diagnosis markers and diagnosis methods for mental illness of children and adolescents, in particular to a plasma metabolic diagnosis marker and a diagnosis system for depression of children and adolescents. The plasma metabolic diagnosis marker comprises at least one of cyclic adenosine monophosphate, methylpyridine, unsaturated undecylenoyl carnitine, histidinol, prolyl hydroxyproline, tryptophan and glucuronic acid. The present application analyzes non-target metabolomics data, and determines seven candidate diagnosis markers for depression of children and adolescents from common differential metabolites. Then, through targeted metabolomics research, diagnosis markers and their combinations with ideal diagnosis effects are determined, and a diagnosis model is constructed. The gender factor is also included in the diagnosis model to further improve the diagnosis effect. The present application can solve the technical problem that the reliability and promotion potential of the plasma metabolic biomarkers for depression of children and adolescents in the prior art are not ideal, and has an ideal promotion and application prospect.

Description

Technical Field

[0001] This invention relates to the technical field of diagnostic biomarkers and methods for mental illness in children and adolescents, specifically to plasma metabolic diagnostic biomarkers and diagnostic systems for depression in children and adolescents. Background Technology

[0002] According to the World Health Organization (WHO), approximately 320 million people worldwide suffer from depression, with about 95 million in China, making depression the second leading cause of disease burden globally. Notably, children and adolescents have become a high-risk group for depression. The first epidemiological report on mental disorders in children and adolescents in my country in 2021 showed a prevalence rate of 17.5%, with depression accounting for approximately 3%. Based on this, it is estimated that about 15 million minors in my country are suffering from depression. Compared to adult depression, childhood and adolescent depression is characterized by difficulties in diagnosis and poor treatment outcomes (only a 22% stable remission rate), leading to 62% of children and adolescents with depression experiencing strong suicidal ideation and / or suicidal behavior. This not only seriously affects the growth, development, and mental and physical health of minors but also places a heavy burden on their families and society as a whole. Furthermore, compared to adult depression, childhood and adolescent depression differs significantly in clinical manifestations, treatment outcomes, and prognosis.

[0003] To address these challenges, a significant amount of research has focused on the diagnosis, treatment, and biological mechanisms of depression in children and adolescents. Among these areas, improving diagnostic accuracy is the most pressing clinical need and the foundation for more efficient research in other areas. To enhance the diagnostic effectiveness of depression in children and adolescents, studies have explored various methods to identify biomarkers, including genomics, proteomics, metabolomics, and the microbiome. Among these omics technologies, metabolomics, used to determine the metabolic profile of vital activities, is one of the best tools for identifying specific disease biomarkers and molecular mechanisms.

[0004] A meta-analysis of peripheral blood metabolomics data from 46 articles in patients with depression suggests dysbiosis in their peripheral blood metabolomics. However, most of the included articles focused on adults, with only three focusing on children and adolescents. Two of these articles on adolescent depression included only 14 cases of depression / 7 healthy controls and 11 cases of depression / 13 healthy controls. Another non-targeted metabolomics study, including 84 cases of depression / 50 healthy controls, was published by the inventors' team in 2018. However, our study had a relatively small sample size, including only 134 participants (84 cases of depression and 50 healthy controls), and the accuracy of the revealed biomarkers requires further investigation, which limits the clinical applicability of the results.

[0005] Therefore, there is an urgent need to further improve the reliability and application potential of plasma metabolic biomarkers for depression in children and adolescents, and to discover new biomarkers or combinations of biomarkers with better accuracy, so as to better achieve effective diagnosis of depression in children and adolescents. Summary of the Invention

[0006] The present invention aims to provide the application of plasma metabolic diagnostic biomarkers in the construction of a diagnostic system for childhood and adolescent depression, in order to solve the technical problems of unsatisfactory reliability and promotion potential of plasma metabolic biomarkers for childhood and adolescent depression in the prior art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The application of plasma metabolic diagnostic markers in constructing a diagnostic system for depression in children and adolescents, wherein the plasma metabolic diagnostic markers include at least one selected from cyclic adenosine monophosphate, methylpyridine, unsaturated undecenoylcarnitine, histidine, prolyl hydroxyproline, tryptophan, and glucuronic acid; the structural formula of unsaturated undecenoylcarnitine is:

[0009]

[0010] Furthermore, the plasma metabolic diagnostic marker is cyclic adenosine monophosphate (cAMP).

[0011] Furthermore, the plasma metabolic diagnostic marker is composed of cyclic adenosine monophosphate and glucuronic acid.

[0012] Furthermore, the plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate (cAMP), unsaturated undecenoylcarnitine, histidine, prolyl hydroxyproline, tryptophan, and glucuronic acid.

[0013] Furthermore, the plasma metabolic diagnostic marker is composed of cyclic adenosine monophosphate, glucuronic acid, and unsaturated undecenoylcarnitine.

[0014] Furthermore, the plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate (cAMP), unsaturated undecenoylcarnitine, histidine, prolyl hydroxyproline, tryptophan, and glucuronic acid.

[0015] This technical solution also provides a diagnostic system for childhood and adolescent depression, characterized in that it includes a data acquisition unit and a diagnostic unit;

[0016] The data acquisition unit includes a content detection subunit, or includes a content detection subunit and a gender information acquisition subunit; the content detection subunit is used to detect the content of plasma metabolic diagnostic markers in plasma; the plasma metabolic diagnostic markers include at least one of cyclic adenosine monophosphate, methylpyridine, unsaturated undecenoylcarnitine, histidine, prolyl hydroxyproline, tryptophan, and glucuronic acid; the structural formula of unsaturated undecenoylcarnitine is:

[0017]

[0018] The diagnostic unit is used to obtain the test variable value P for the risk of diagnosing depression through a logistic regression model.

[0019] The logistic regression model is: Logit(P)=a1×M1+a2×M2+a3×M3+a4×M4+a5×M5+a6×M6+a7×M7+a8×M8+b;

[0020] Among them, M1-M7 are independent variables, which are used to input the numerical values ​​of the plasma content of cyclic adenosine monophosphate, glucuronic acid, unsaturated undecenoylcarnitine, methylpyridine, histidine, prolyl hydroxyproline and tryptophan, respectively; among the independent variables of M1-M7, at least one independent variable is input with the numerical value of the corresponding metabolite content, and the remaining independent variables are set to 0.

[0021] M8 is the independent variable. When considering gender factors, M8 is used to input gender information, with a value of 1 for males and 2 for females; when not considering gender factors, M8 is set to 0.

[0022] a1-a8 are the coefficients of M1-M8 respectively, and b is a constant;

[0023] 0≤P≤1, the diagnosis is achieved by comparing the P value with the critical value.

[0024] Furthermore, the logistic regression model is constructed using the following method: collecting information on the content of plasma metabolic diagnostic markers in plasma from several samples, or simultaneously collecting information on the content of plasma metabolic diagnostic markers in plasma and gender information from several samples, thereby obtaining a dataset; using the logistic regression algorithm to train the model on the dataset to obtain the logistic regression model;

[0025] The critical value is obtained by fitting an ROC curve using a logistic regression model in the dataset. The value of the test variable corresponding to the maximum value of the Youden index of the ROC curve is the critical value used for diagnosis.

[0026] Furthermore, plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate, glucuronic acid, and unsaturated undecenoylcarnitine:

[0027] Without considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula:

[0028] Logit(P)=-0.2102011×M1+0.0024516×M2-0.0087530×M3+2.5230796.

[0029] Alternatively, considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula:

[0030] Logit(P)=-0.219851×M1+0.0024516×M2-0.011070×M3-0.671532M8+4.026209.

[0031] Furthermore, the plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate, glucuronic acid, unsaturated undecenoylcarnitine, methylpyridine, histidine, prolyl hydroxyproline, and tryptophan:

[0032] Without considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula:

[0033] Logit(P)=-0.2722×M1+0.002628×M2-0.01476×M3+0.002021×M4+0.0005301×M5+0.000555×M

[0034] 6 + 0.000008763 × M7 - 0.8001;

[0035] Alternatively, considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula:

[0036] Logit(P)=-0.2756×M1+0.002628×M2-0.001359×M3+0.001737×M4+0.0006020×M5+0.0007109

[0037] ×M6+0.000009975×M7+0.5543×M8-2.336.

[0038] Furthermore, the content detection subunit includes a high-performance liquid chromatography-mass spectrometry (HPLC-MS) device.

[0039] Furthermore, the high-performance liquid chromatography (HPLC) conditions in the HPLC-MS / MS equipment are as follows:

[0040] Chromatographic column: UPLC BEH Amide column; Mobile phase A: aqueous solution of CH3COONH4 and NH4OH; Mobile phase B: acetonitrile; Flow rate: 0.5 mL / min; Sample loading volume: 4 μL; Gradient elution mode was used.

[0041] Furthermore, the gradient elution program is as follows: 0-0.5 min, 95% mobile phase B; 0.5-7 min, 95%-65% mobile phase B; 7-8 min, 65%-40% mobile phase B; 8-9 min, 40% mobile phase B; 9-9.1 min, 40%-95% mobile phase B; 9.1-12 min, 95% mobile phase B.

[0042] Furthermore, the mass spectrometry conditions in the high performance liquid chromatography-mass spectrometry (HPLC-MS) equipment are: ESI positive ion and negative ion modes; the ESI source conditions are set as follows: sheath gas temperature 350℃; sheath gas flow rate 12L / min; capillary voltage: +3000V, -2500V.

[0043] Furthermore, for cAMP, the precursor ion is 330.06 Da, and the daughter ions are 136.00 Da and 119.00 Da; for unsaturated undecenoic acid, the precursor ion is 328.25 Da, and the daughter ions are 85.00 Da and 57.10 Da; for glucuronic acid, the precursor ion is 193.03 Da, and the daughter ions are 113.10 Da and 131.10 Da; for histidine, the precursor ion is 124.09 Da, and the daughter ions are 81.10 Da and 95.00 Da; for tryptophan, the precursor ion is 205.10 Da, and the daughter ions are 146.00 Da and 188.00 Da; for prolyl hydroxyproline, the precursor ion is 229.12 Da, and the daughter ions are 70.10 Da and 131.90 Da; for methylpyridine, the precursor ion is 94.07 Da, and the daughter ions are 50.10 Da and 52.10 Da.

[0044] In summary, the principle and beneficial effects of this technical solution are as follows:

[0045] This technical solution explores plasma metabolite changes in childhood and adolescent depression by combining untargeted and targeted metabolomics, identifying plasma biomarkers that can be used for the diagnosis of childhood and adolescent depression and validating them in both untargeted and targeted cohorts. The untargeted cohort includes three sets: an untargeted training set, an untargeted validation set, and an untargeted test set. The targeted cohort includes two sets: a targeted training set and a targeted prediction set.

[0046] In this study, batch calibration was first performed on non-target metabolomics data from three batches of 1073 samples across the non-target training set, non-target validation set, and non-target test set. Then, a combined univariate metabolomics analysis was conducted on both the non-target training set and the non-target validation set. One hundred differentially expressed metabolites were identified. Subsequently, based on the non-target discovery set (70% of the non-target training set) and the non-target internal validation set (30% of the non-target training set) divided from the non-target training set, feature screening was conducted using genetic algorithms and cross-validation, while considering collinearity and substitutability. Further feature screening identified seven candidate plasma metabolic diagnostic biomarkers for childhood and adolescent depression: cyclic adenosine monophosphate (cAMP), picoline, undecenoylcarnitine (Car(11:1), histidine), prolylhydroxyproline, tryptophan, and glucuronic acid. Then, these seven metabolites were detected in a new targeted cohort containing 400 multicenter samples using targeted metabolomics. Finally, three metabolites, namely cyclic adenosine monophosphate (cAMP), undecenoylcarnitine (Car(11:1), undecenoylcarnitine), and glucuronic acid, passed the targeted validation and were used as the final diagnostic biomarkers.

[0047] A diagnostic model equation was constructed based on three final biomarkers, and its diagnostic efficacy was evaluated in both targeted and non-targeted cohorts. The combined diagnostic model based on the three biomarkers demonstrated good diagnostic efficacy. The optimal cut-off value was then calculated using YoudenIndex, which showed good diagnostic performance. Subsequently, the correlation between the three final diagnostic biomarkers and clinical indicators was evaluated, revealing a stable correlation between sex and unsaturated undecylcarnitine (Car(11:1)). Subgroup analysis also showed that the enrolled population may influence carnitine concentration in depressed patients. Therefore, we fitted specific cut-off values ​​to different sex subgroups and enrolled population subgroups, finding that subgroup-specific cut-off values ​​based on the same diagnostic model better met the diagnostic needs of different populations compared to the global cut-off value. It is important to clarify that our subgroup-specific cut-off value results should be correctly interpreted as follows: our results suggest that each region and laboratory should fit its own normal reference levels (specific cut-off values) for men and women based on its own test subjects, and should not be interpreted as simply copying our specific cut-off values ​​fitted based on a limited sample subgroup. If necessary, our cut-off values ​​can be used temporarily at the beginning of the project, and then fitted to the normal reference values ​​for the region after the local testing database is established.

[0048] Car(11:1) is an unsaturated acylcarnitine and an intermediate product of unsaturated fatty acid β-oxidation metabolism. Our results suggest that carnitine (Car(11:1)) levels are affected by sex; plasma carnitine (Car(11:1)) concentrations in normal female children and adolescents are lower than in normal males, and depressive patients also show a decrease in carnitine (Car(11:1)) concentrations compared to normal individuals. This suggests that the higher incidence of depression in women may involve naturally lower plasma carnitine (Car(11:1)) concentrations in women, and this sex difference results in women receiving less carnitine protection in depression.

[0049] Glucuronic acid, a plasma metabolic diagnostic marker, is a substance in the uronic acid pathway, a bypass pathway in glucose metabolism, and is widely expressed in vivo. On the one hand, its product, UDP-glucuronic acid, has a detoxification function and can also be used in mucopolysaccharide synthesis. On the other hand, the uronic acid pathway is linked to the pentose phosphate pathway, forming a bypass portion of the glucose metabolism network. Currently, research on glucuronic acid in depression is limited. Our observation of glucuronic acid upregulation in the plasma of children and adolescents may be due to the following reasons: depression involves energy impairment, leading to impaired aerobic oxidation of glucose (downregulation of tricarboxylic acid cycle metabolites in pathway analysis), resulting in compensatory upregulation of bypass pathways such as glycosides and pentose phosphate. Furthermore, antidepressant treatment can also secondaryly inhibit the activity of pentose phosphate-related enzymes, leading to upregulation of glucuronic acid, which is associated with the pentose phosphate pathway.

[0050] cAMP exists in trace amounts within cells, acting as an intracellular second messenger. It is formed by the cyclization of ATP catalyzed by adenylate cyclase (AC), activated by first messenger hormones or other molecular signals. cAMP is ultimately inactivated by phosphodiesterase (PDE) to 5'-AMP. AC and PDE regulate intracellular cAMP concentration in two different ways, thus affecting the function of cells, tissues, and organs. When AC activity increases, cAMP concentration increases; when PDE concentration increases, cAMP concentration decreases. cAMP participates in synaptic transmission in nerve cells. After the receptors on the postsynaptic membrane are activated by neurotransmitters, AC is activated, forming cAMP, which in turn activates PKA, leading to phosphorylation of membrane proteins and regulation of ion channels and gene expression.

[0051] Furthermore, differential metabolite and pathway analyses revealed significant abnormalities in amino acid metabolism among plasma diagnostic biomarkers. Although these differences were unstable in multicenter studies, amino acids are less stable as diagnostic biomarkers than carnitine (Car(11:1)), glucuronic acid, and cAMP, possibly due to increased heterogeneity from multicenter studies. Alterations in amino acid metabolic pathways in childhood, adolescents, and adults with depression have been widely reported in the literature, but consistency among these reports has been poor.

[0052] In summary, the advantages of this technical solution are as follows: Existing technologies for biomarkers of depression in children and adolescents generally employ non-targeted metabolomics detection, resulting in a lack of inter-batch validation and, moreover, absolute quantitative targeted validation of the obtained diagnostic biomarkers. While non-targeted metabolomics exhibits significant inter-batch variability, this technical solution creatively solves this problem through standardized detection methods and inter-batch algorithm correction, making the three batches of non-targeted metabolites comparable. Based on the inter-batch correction of the three batches of non-targeted metabolites, this technical solution uses a genetic algorithm for feature selection and also incorporates collinearity and model performance to select metabolites. Finally, targeted metabolomics validation is performed. Targeted validation can clarify the objective absolute concentration of metabolites, making the diagnostic formula more reliable and enabling direct clinical application. The combination of carnitine (Car(11:1)), glucuronic acid, and cAMP diagnostic biomarkers used in this technical solution is the optimal biomarker for diagnosis, and the diagnostic model constructed from it can accurately diagnose depression. Attached Figure Description

[0053] Figure 1 For the research process flowchart.

[0054] Figure 2 For screening differentially expressed metabolites in the non-target validation set of Example 2, which is for depression and normal groups (A: Metabolite volcano plots of the MDD and HC groups in the training set; B: Metabolite volcano plots of the MDD and HC groups in the validation set; C: Venn plot showing the union of differentially expressed metabolites from the training and validation sets).

[0055] Figure 3 Metabolic characteristics of the depressed and normal groups in the non-target cohort of Example 2 (downregulated metabolites are in blue, and upregulated metabolites are in red).

[0056] Figure 4 This is a simplified schematic diagram of the differential metabolic pathways between the depressed and normal groups in the non-target cohort of Example 2 (downregulated metabolites are in blue, and upregulated metabolites are in red).

[0057] Figure 5 The differential distribution of diagnostic biomarkers in Example 2 across the non-target training set, non-target validation set, and non-target test set.

[0058] Figure 6For Example 2, the non-target metabolic diagnostic assessment of seven candidate diagnostic biomarkers for depression was performed between normal and normal individuals (A: combined diagnosis of seven metabolites, AUC value of ROC curve for MDD-HC in the targeted training set was 0.881; AUC value of ROC curve for MDD-HC in the targeted training set was 0.838; B: combined diagnosis of seven metabolites, AUC value of ROC curve for MDD-HC in the discovery set of the non-target training set was 0.941; AUC value of ROC curve for MDD-HC in the internal validation set of the non-target training set was 0.919; C: combined diagnosis of seven metabolites, AUC value of ROC curve for MDD-HC in the non-target validation set was 0.810; AUC value of ROC curve for MDD-HC in the non-target test set was 0.883).

[0059] Figure 7 This is a targeted metabolic assessment of diagnostic biomarkers between depression and normality in Example 3.

[0060] Figure 8 For Example 3, the targeted and non-target metabolic diagnostic assessment of three final diagnostic biomarkers for depression was conducted between normal and normal individuals (A: Combined diagnosis of the three diagnostic metabolites: AUC value of ROC curve for MDD-HC was 0.849 in the targeted training set and 0.858 in the targeted prediction set; B: Combined diagnosis of the three diagnostic metabolites: AUC value of ROC curve for MDD-HC was 0.878 in the discovery set of the non-targeted training set and 0.863 in the internal validation set of the non-targeted training set; C: Combined diagnosis of the three diagnostic metabolites: AUC value of ROC curve for MDD-HC was 0.793 in the non-targeted validation set and 0.833 in the non-targeted test set).

[0061] Figure 9 Correlation analysis of the three final diagnostic biomarkers and clinical information in Example 3 (A: Correlation analysis of the non-target training set; B: Correlation analysis of the non-target validation set; C: Correlation analysis of the non-target test set; D: Correlation analysis of the target test set).

[0062] Figure 10 Sex subgroup analysis of carnitine Car (11:1) in Example 3 (A: Correlation analysis of non-target training set; B: Correlation analysis of non-target validation set; C: Correlation analysis of non-target test set; D: Correlation analysis of target test set).

[0063] Figure 11Sex subgroup analysis of carnitine Car (11:1) in different centers in Example 3 (A: Correlation analysis of non-target set in Chong Qing center; B: Correlation analysis of target set in Chong Qing center; C: Correlation analysis of target set in Chang Sha center; D: Correlation analysis of target set in Nan Jing center). Detailed Implementation

[0064] The present invention will be further described in detail below with reference to embodiments, but the implementation of the present invention is not limited thereto. Unless otherwise specified, the technical means used in the following embodiments and experimental examples are conventional means well known to those skilled in the art, and the materials and reagents used can all be obtained commercially.

[0065] To further improve the reliability and application potential of plasma metabolic biomarkers for childhood and adolescent depression, this study explored changes in plasma metabolites in children and adolescents with depression by combining untargeted and targeted metabolomics. Plasma biomarkers suitable for the diagnosis of childhood and adolescent depression were identified and validated in both untargeted and targeted cohorts. The untargeted cohort included three sets: an untargeted training set, an untargeted validation set, and an untargeted test set. The targeted cohort included two sets: a targeted training set and a targeted prediction set. See the detailed research flowchart below. Figure 1 .

[0066] Example 1:

[0067] (1) Patient enrollment and sample collection

[0068] Research flowchart as follows Figure 1 As shown in Figure A, all children and adolescents with MDD undergoing non-targeted metabolomics testing were recruited from the Department of Psychiatry at the First Affiliated Hospital of Chongqing Medical University. A healthy control group matched for patient demographics was also recruited from the First Affiliated Hospital of Chongqing Medical University through advertising. Simultaneously, healthy controls were recruited from two communities in the main urban area of ​​Chongqing and Dianjiang District.

[0069] Patients with myelopathic disease (MDD) in children and adolescents were recruited from the Department of Psychiatry at the First Affiliated Hospital of Chongqing Medical University, the Department of Psychiatry at Xiangya Hospital in Changsha, and Nanjing Brain Hospital. Healthy controls matched to the patients' demographic information were recruited through advertising from the First Affiliated Hospital of Chongqing Medical University and Xiangya Hospital in Changsha. Simultaneously, healthy controls were also recruited from two communities in Dianjiang and Bishan districts of Chongqing.

[0070] The diagnosis of depression in all children and adolescents was confirmed by two experienced psychiatrists according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V). Participants were included in the study only if they met the following criteria: 1) were aged between 12 and 18 years, 2) were diagnosed with MDD or were healthy individuals matched to the depression group, and 3) provided informed consent from both themselves and their legal guardians. Individuals are excluded if they meet any of the following criteria: 1) a history of head injury resulting in persistent loss of consciousness or cognitive consequences; 2) a current or lifelong diagnosis of a serious neurological disorder or other mental disorder, such as epilepsy, encephalitis, autism, bipolar disorder, schizophrenia, attention-deficit / hyperactivity disorder, or obsessive-compulsive disorder; 3) the presence of a chronic physical illness that may significantly affect peripheral metabolism or neurological function, such as hepatitis, chronic nephrosis, or chronic enteritis; or 4) a history of substance abuse or dependence.

[0071] Then, general demographic and clinical characteristics, including sex, age, body mass index (BMI), and symptom scales, were collected. Patients with HCs and MDD were clinically assessed using the 24-item Hamilton Depression Rating Scale (HAMD-24) and the 14-item Hamilton Anxiety Rating Scale (HAMA-14). All participants in the non-target cohort were divided into a non-target training set (Tr, n = 563), a non-target validation set (Va, n = 397), and a non-target test set (Te, n = 113). The non-target training set was randomly divided into a discovery set (Di, n = 394) and an internal validation set (IV, n = 179) at a ratio of 70% to 30%. All participants in the targeted cohort (Ta: targeted set, n=400) were divided into a targeted training set (TT: targeted training set, n=200) and a targeted prediction set (TP: targeted prediction set, n=200). Participants and their legal guardians read and signed informed consent forms before participation. This research protocol has been approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University.

[0072] All blood samples were collected from participants via venipuncture between 8:00 AM and 12:00 PM and placed in tubes coated with EDTA (anticoagulant). The tubes were then centrifuged at 3000 rpm and 4°C for 10 minutes. After centrifugation, the plasma was separated and stored at -80°C until transported to the metabolomics laboratory on dry ice for metabolomics analysis.

[0073] (2) Demographic characteristics and clinical information of the subjects

[0074] Detailed demographic and clinical characteristics of the recruited non-target cohort participants are shown in Table 1. A total of 1073 children and adolescents were recruited, including 509 patients with MDD and 564 healthy controls. The non-target discovery set included 256 patients with depression and 307 healthy controls; the non-target validation set included 197 patients with depression and 200 healthy controls; and the non-target test set included 56 patients with depression and 57 healthy controls. The participants' ages ranged from 10 to 18 years. The mean ± SD for depression and normality in the non-target discovery set were 15.4 ± 1.7 and 15.4 ± 1.7, respectively; the mean ± SD for depression and normality in the non-target validation set were 15.0 ± 1.8 and 14.6 ± 1.7, respectively; and the mean ± SD for depression and normality in the non-target test set were 15.2 ± 1.4 and 14.7 ± 1.5, respectively. In all three cohorts, there were no significant differences between the depression group and the normal control group in age (non-target discovery set P = 0.178, non-target validation set P = 0.178, non-target test set P = 0.178), sex (non-target discovery set P = 0.178, non-target validation set P = 0.178, non-target test set P = 0.178), or BMI (non-target discovery set P = 0.178, non-target validation set P = 0.178, non-target test set P = 0.178).

[0075] Table 1: Demographic characteristics of depressed individuals and healthy controls in the non-target cohort

[0076]

[0077] In the table above, continuous variables are expressed as mean ± SD. Gender was determined using a chi-square test, while age, BMI, HAMA, and HAMD were determined using a Kruskal-Wallis test. The abbreviations are: MDD (major depressive disorder); HC (health control); BMI (body mass index); HAMD (24-item Hamilton Depression Rating Scale); and HAMA (14-item Hamilton Anxiety Rating Scale). Subscript 'a' indicates a chi-square test, and subscript 'b' indicates a Kruskal-Wallis test.

[0078] Detailed demographic and clinical characteristics of the recruited participants in the targeted cohort are shown in Table 2. A total of 400 children and adolescents were recruited, including 200 patients with MDD and 200 healthy controls. The ChongQing center included 84 patients with depression and 138 healthy controls; the Chang Sha center included 66 patients with depression and 62 healthy controls; and the NanJing center included 50 patients with depression. Participants ranged in age from 11 to 18 years. In all three scenarios, there were no significant differences between the depression group and the normal control group in age (targeted training set P = 0.261, targeted prediction set P = 0.727, all subjects (Targeted Set) P = 0.307), sex (targeted training set P = 0.884, targeted prediction set P = 0.767, all subjects (Targeted Set) P = 0.917), or BMI (targeted training set P = 0.906, targeted prediction set P = 0.263, all subjects (Targeted Set) P = 0.422).

[0079] Table 2: Demographic characteristics of depressed and healthy controls in the targeted cohort

[0080]

[0081] Example 2: Non-targeted metabolomics research

[0082] (1) Non-targeted metabolomics detection and raw data processing

[0083] (1.1) Preparation of main reagents and samples for LC-MS analysis

[0084] Non-target metabolism experimental procedures are as follows Figure 1 As shown in B. LC-MS (Liquid Chromatography-Mass Spectrometry) grade water (H2O) and methanol (MeOH) were purchased from Honeywell (Muskegon, USA). Ammonium hydroxide (NH4OH) and ammonium acetate (NH4OAc) were purchased from Sigma-Aldrich (St. Louis, USA). Metabolite chemical standards were purchased from J&K (Beijing, China), Sigma (St. Louis, USA), Carbosynth (Berkshire, UK), TCI (Tokyo, Japan), and Energy Chemical (Shanghai, China).

[0085] Human plasma samples (50 μL) were extracted using 150 μL MeOH and internal standards (d3-leucine and d6-phenylalanine). The samples were then shaken for 30 seconds and sonicated for 15 minutes. To precipitate proteins, the samples were incubated at -20 °C for 1 hour, followed by centrifugation at 13,500 rpm and 4 °C for 15 minutes. The resulting supernatant was transferred to a high-performance liquid chromatography (HPLC) vial and stored at -80 °C for LC-MS / MS analysis.

[0086] (1.2) LC-MS detection

[0087] LC-MS analysis methods are described in the literature (WANG H, JIA H. Serum metabolic traits reveal therapeutic toxicities and responses of neoadjuvant chemoradiotherapy inpatients with rectal cancer[J]. 2022, 13(1):7802.). Data acquisition was performed using a Thermo Scientific Vanquish UHPLC system coupled to a Thermo Scientific Orbitrap Exploris 480. LC separation was performed using a Waters ACQUITY UPLC BEH Amide column (particle size 1.7 μm; length 100 mm × inner diameter 2.1 mm) and a Kinetex C18 column (2.6 μm, 2.1 × 100 mm), with the column temperature maintained at 25 °C. For HILIC analysis, the mobile phases for positive (ESI+) and negative (ESI-) modes were: A was 25 mM ammonium hydroxide (NH4OH) + 25 mM ammonium acetate (NH4OAc) in water, and B was acetonitrile (ACN). The flow rate was 0.5 mL / min, with the following gradient settings: 0–0.5 min, 95% B; 0.5–7 min, 95% B to 65% B; 7–8 min, 65% B to 40% B; 8–9 min, 40% B; 9–9.1 min, 40% B to 95% B; 9.1–12 min, 95% B. The injection volume was 2 μL. For reversed-phase liquid chromatography (RPLC) analysis, mobile phase A was 0.01% acetic acid in water, and mobile phase B was a mixture of IPA and ACN (1:1), using both positive (ESI+) and negative (ESI-) modes. The flow rate was 0.3 mL / min, with the following gradient settings: 0–1 min: 1% B; 1–8 min: 99% B; 8–9 min: 99% B; 9.0–9.1 min, 99% B to 1% B; 9.1–12 min: 1% B. The injection volume was 2 μL. All samples were randomly injected during data acquisition. Data acquisition was performed in full MS scan mode with positive / negative ion polarity switching. Information-dependent acquisition (IDA) mode was used for quality control (QC) samples to obtain MS / MS spectra. Source parameters were set as follows: spray voltage 3,000 V in positive ion mode and -3,000 V in negative ion mode. Assist gas heater temperature was set to 350 °C. Substrate gas was set to 50 alb. Assist gas was set to 15 alb. Capillary temperature was set to 400 °C. Full MS resolution was set to 60,000, and the AGC target in positive ion mode was set to 1e6. Maximum IT was set to 100 ms. Mass range was set to 70–1,200 Da.For dd-MS2 settings, the MS resolution was set to 30,000, and the AGC target was set to 1e5. The maximum IT was set to 60 ms. The Top N was set to 6. The isolation width was set to 1.0 m / z Da. MS / MS spectra of QC samples were obtained at SNCE 20-30-40%. Dynamic exclusion was set to 3.0 seconds, and isotope exclusion was enabled.

[0088] (1.3) Metabolomics Data Processing

[0089] Data processing methods are described in the references (SHEN X, WANG R, XIONG X, et al. Metabolic reaction network-based recursive metabolite annotation for untargeted metabolomics[J]. 2019, 10(1): 1516.; ZHOU Z, LUO M, ZHANG H, et al. Metabolite annotation from knowns to unknowns through knowledge-guided multi-layer metabolic networking[J]. 2022, 13(1): 6656.). In short, the raw MS data (.raw) files were first converted to mzXML format using ProteoWizard (version 3.0.20360). Then, peak detection, retention time correction, and peak alignment were performed on the sample mzXML data files using the R package "xcms" (version 3.2; https: / / bioconductor.org / packages / release / bioc / html / xcms.html). Key parameter settings were as follows: Method, “centWave”; ppm, 10; snhr, 3; peakwidth, c(5,30); minfrac, 0.5. Then, missing value imputation and data normalization were performed using MetFlow software (http: / / metflow.zhulab.cn / ). In short, missing values ​​were imputed using the k-nearest neighbor algorithm with the 10 nearest neighbors, and then the data was normalized using a support vector algorithm based on QC samples to eliminate unwanted systematic errors occurring in batches. Metabolic peaks with an RSD less than 30% in QC samples were used for subsequent analysis. Metabolite annotation was performed using MetDNA (version 1.2.2; http: / / metdna.zhulab.cn / ). Metabolite annotation parameters were set to “HILIC” or “RP” depending on the HPLC mode, and the collision energy was set to “30” or “SNCE_20_30_40%”. We performed metabolite annotation in both positive and negative modes. According to the Metabolomics Standards Initiative (MSI) definition, we annotate metabolites by matching MS1, RT, and MS / MS spectra with our laboratory's metabolite library, and by matching MS1 and MS / MS2 spectra with public metabolite libraries (primarily from NIST 2017).

[0090] (1.4) Batch effect processing of non-targeted metabolomics data

[0091] Non-target metabolism experimental procedures are as follows Figure 1 As shown in B, we only retained metabolites detected simultaneously in both the training and validation sets. We then aligned these two batches of metabolites based on their liquid chromatographic mode and ionization polarity. Next, we used a combat algorithm to correct the intensity difference between the two datasets to reduce batch effects. After correction, metabolites with differences exceeding 40% were removed, resulting in an integrated metabolomics dataset containing 516 metabolites. Finally, in the test set, we also aligned the metabolites with the training set based on their liquid chromatographic mode and ionization polarity, and used a combat algorithm to correct the intensity difference between the two datasets, resulting in an integrated metabolomics dataset containing 470 metabolites.

[0092] (2) Identify differential metabolites and metabolic pathways between the depressed and normal groups.

[0093] First, we used the Wilcoxon-Mann-Whitney test on both the training and validation sets to determine the differential levels of all metabolites in the comparison between depression and HC. Furthermore, we calculated the fold change (FC) of each metabolite based on its ratio to the mean. Metabolites with a Wilcoxon-Mann-Whitney test p < 0.05 on both the training and validation sets, and exhibiting the same trend of change, were selected as differentially expressed metabolites in plasma for childhood and adolescent depression. Finally, to further investigate specific metabolic pathways in childhood and adolescent depression, we annotated the expressed metabolites according to the Kyoto Encyclopedia of Genetics and Genomes (KEGG) (https: / / www.kegg.jp / ) and drew simplified pathway diagrams for the differentially expressed metabolites.

[0094] (3) Screening and evaluation of normal diagnostic criteria for depression

[0095] (3.1) Screening of normal diagnostic criteria for depression

[0096] To further screen for diagnostic biomarkers from differentially expressed plasma metabolites in children and adolescents with depression, we randomly split the training set into a discovery set of 394 samples (70% of the training set) and an internal validation set of 169 samples (30% of the training set) for cross-validation. First, we removed metabolites with low signal intensity (<1×10⁻⁶) from both the training and validation sets. 6Then, based on the "deap" data package in Python, a genetic algorithm was used for feature selection. We performed selection, crossover, and mutation operations on an initial set of features randomly generated by the genetic algorithm to generate new feature combinations. We retained combinations with high fitness and removed those with low fitness. Fitness values ​​were calculated using logistic regression, and 3-fold cross-validation was performed on the discovery set to obtain robust fitness values. We repeated the selection process 40 times (feature count = 10), and then obtained 13 candidates based on the criterion of being selected more than 10 times. Additionally, glucuronic acid was retained due to its good discriminative performance, and cAMP was also screened and retained from low-abundance substances due to its good discriminative performance and potential neurobiological functions.

[0097] Among the 15 candidates, based on the model's discriminative performance and the collinearity and substitutability of the metabolites (Fisher's exact ptest), we selected seven metabolites as candidate diagnostic biomarkers: cyclic adenosine monophosphate (cAMP), picoline, undecenoylcarnitine (Car(11:1), undecenoylcarnitine), histidine, prolylhydroxyproline, tryptophan, and glucuronic acid.

[0098] (3.2) Non-target assessment of seven normal alternative diagnostic biomarkers for depression

[0099] We also evaluated the differences in the seven diagnostic biomarkers between childhood and adolescent depression and HC using the Wilcoxon-Mann-Whitney test on the test set. We visualized the relative concentrations of these seven metabolites in the training, validation, and test sets using box plots with the "ggplot2" function in R. Subsequently, we constructed a logistic regression model based on the seven potential biomarkers using the "glm" function in R, utilizing the discovery set. We then evaluated the diagnostic power of the logistic regression model based on the seven candidate biomarkers using receiver operating characteristic (ROC) curves on the discovery set (70% of the training set), the internal validation set (70% of the training set), the validation set, and the test set.

[0100] (4) Differences in metabolites and metabolic pathways between the depressed and normal groups

[0101] First, based on the fold change (FC) of metabolites and the p-value of the Wilcoxon-Mann-Whitney test, we identified 333 differentially expressed metabolites related to childhood and adolescent depression from the non-target training set, of which 51 were upregulated and 282 were downregulated. Figure 2 A). Subsequently, using the same method, we identified 241 differentially expressed metabolites in the non-target validation set, of which 67 were upregulated and 174 were downregulated. Figure 2 B). Ultimately, 160 metabolites were found to differ between the two sets, and 100 of them showed the same trend, with 5 upregulated and 95 downregulated. Figure 2 C).

[0102] We then used heatmaps to visualize the expression of the 100 selected metabolites in the non-target discovery and non-target validation sets, and labeled the major pathway categories to which the metabolites belonged on the heatmaps. Among them, 52 metabolites were enriched in the lipid metabolism pathway, followed by 25 metabolites in the amino acid metabolism pathway, 4 metabolites in the nucleotide metabolism pathway, 4 metabolites in the cofactor and vitamin metabolism pathway, 3 metabolites in the carbohydrate metabolism pathway, and 12 metabolites in other metabolic pathways. Figure 3 Furthermore, our simplified pathway diagrams based on differentially metabolized metabolites also show that the vast majority of metabolites in lipid metabolism, amino acid metabolism, nucleotide metabolism, cofactor and vitamin metabolism, carbohydrate metabolism, and other metabolic pathways are downregulated. Interestingly, the major tricarboxylic acid cycle pathway in carbohydrate metabolism is downregulated, while the alternative pentose pathway is upregulated. Figure 4 ).

[0103] (5) Identification of candidate diagnostic biomarkers

[0104] We used a genetic algorithm and cross-validation, considering both collinearity and substitutability, to identify cyclic adenosine monophosphate (cAMP, chemical formula C10) from plasma metabolites. 10 H 12 N5O6P), methylpyridine (picoline, chemical formula C6H7N), undecenoylcarnitine (Car(11:1), chemical formula C 18 H 33 NO4), histidine (chemical formula C6H4) 11 N3O), prolylhydroxyproline (chemical formula C3O), prolylhydroxyproline (chemical formula C3O) 10 H 16 N2O4), tryptophan (chemical formula C2O4), 11 H 12N2O2), glucuronic acid (chemical formula C6H2O2), 10 Seven candidate diagnostic biomarkers were identified (O7). Glucuronate was upregulated in depression; cyclic adenosine monophosphate (cAMP), methylpyridine, unsaturated undecenoylcarnitine (Carb(11:1), histidine), prolyl hydroxyproline, and tryptophan were downregulated. Except for prolyl hydroxyproline, which showed no difference in the non-target test set (P = 0.663), the other metabolites showed significant differences in all three cohorts. Figure 5 Histidine, prolyl hydroxyproline, and tryptophan originate from the amino acid metabolism pathway; unsaturated undecenoylcarnitine (Car (11:1)) originates from the fatty acid metabolism pathway; glucuronic acid originates from the glucose metabolism pathway; cyclic adenosine monophosphate (cAMP) originates from the nucleotide metabolism pathway; and methylpyridine originates from other pathways. Figure 3 , Figure 4 ). Detailed information on these diagnostic biomarkers can be found in Table 3. Among them, the structural formula of unsaturated undecenoylcarnitine is:

[0105]

[0106] Table 3: Metabolic Information for Depression Diagnosis in Non-Target Screening

[0107]

[0108] In the table above, TPvalue refers to the p-value of the training set passing the Kruskal-Wallis test; T.FC is the fold change of the training set; VPvalue refers to the p-value of the validation set passing the Kruskal-Wallis test; and V.FC is the fold change of the validation set.

[0109] (6) Non-target metabolic assessment of seven candidate diagnostic biomarkers

[0110] Our logistic regression model, constructed based on the non-target discovery set and incorporating seven potential biomarkers, achieved good diagnostic efficacy across the non-target discovery set (70% of the non-target training set), the internal non-target validation set (30% of the non-target training set), the non-target validation set, and the non-target test set. Specifically, the AUC of the non-target discovery set ROC was 0.941, the internal non-target validation set ROC was 0.919, the non-target validation set ROC was 0.810, and the non-target test set ROC was 0.883. Figure 6 AB).

[0111] Example 3: Targeted Metabolomics Research

[0112] (1) Preparation of main reagents and samples for targeted metabolism detection

[0113] Targeted metabolism experimental procedure as follows Figure 1As shown in C. The reagents required for the experiment, methanol, ultrapure water, and ammonia, were purchased from Honeywell (LC-MS grade); acetonitrile was purchased from Merk (LC-MS grade); ammonia and ammonium acetate were LC-MS grade. The standards used included Picoline, 3-Picoline (D7, 98%), [13C6]-L-Histidine, Histidinol, Tryptophan, L-Tryptophan (13C11, 99%), Proylhydroxyproline, L-Proline ((2,5,5-D3), Glucuronicacid, [U-13C6]-D-glucuronic acid, L-Carnitine (D9) (Car(12:1)), Undecenoylcarnitine (Car(11:1)), Car(10:1), cAMP, and cAMP_13C5.

[0114] Take 50 μL of plasma sample and place it in a 1.5 mL EP tube. Add 150 μL of ice-cold methanol containing internal standard and vortex for 30 s. Store at -20℃ for 1 h to precipitate protein. Centrifuge at 17000 g and 4℃ for 15 min. Take at least 150 μL of supernatant into an LC-MS vial. The concentrations of internal standards in the sample are as follows: 3-Picoline (D7, 98%) 2 ng / mL, L-TRYPTOPHAN (13C11) 5000 ng / mL, L-Picoline ((2,5,5-D3)) 10 ng / mL, [U-13C6]-D-glucuronic acid 50 ng / mL, cAMP_13C5 2 ng / mL, [13C6]-L-Histidine 200 ng / mL, and L-Carnitine (D9)(Car(12:1)) 2 ng / mL.

[0115] (2) Targeted metabolomics detection

[0116] LC-MS analysis was performed using an HPLC system (1290 series, Agilent Technologies) and a triple quadrupole mass spectrometer (Agilent 6495QqQ, Agilent Technologies). A Waters, UPLC BEH Amide column (1.7 μm, 2.1 × 100 mm) was selected. The mobile phase consisted of 100% H₂O + 25 mM CH₃COONH₄ + 25 mM NH₄OH, and B = 100% acetonitrile, used for both ESI positive and negative ion modes. The linear gradients were 95% B (0–0.5 min), 95%–65% B (0.5–7 min), 65%–40% B (7–8 min), 40% B (8–9 min), 40%–95% B (9–9.1 min), and 95% B (9.1–12 min). The flow rate was 0.5 mL / min, and the sample injection volume was 4 μL. The ESI source conditions are set as follows: sheath gas temp, 350℃; sheath gas flow, 12L / min; capillary voltage positive 3000V, negative 2500V. Table 4 shows the information on various metabolites produced by the MRM method.

[0117] Table 4: Information on various metabolites obtained by the MRM method

[0118]

[0119] (3) Targeted metabolomics data processing

[0120] Skyline (version 22.2.0.351) was used for manual integration of metabolite and internal standard peak areas. First, the raw data file (.d) and the target metabolite's procursor m / z, product m / z, precursor charge, product charge, and RT were imported into Skyline. The target metabolite was manually checked, and the integration interval was adjusted. The peak areas of the internal and external standards were exported, and the ratio of the external to internal standard peak areas was calculated to establish a standard curve. For data collected each day, two sets of standard curves were collected at the beginning and end of each day to establish the standard curve for that day and calculate the content in the quantitative sample. Finally, the sample quantification results were standardized using the laboratory-developed R package metflowR (version 0.99.01).

[0121] (4) Multicenter targeted metabolomics differential analysis and final diagnostic model evaluation

[0122] (4.1) Differential analysis of targeted metabolomics of alternative diagnostic standards

[0123] Targeted metabolic analysis workflow as follows Figure 1 As shown in Figure C. We evaluated the differences in seven diagnostic biomarkers between childhood and adolescent depression and HC using the Wilcoxon-Mann-Whitney test in a multicenter targeted cohort. The absolute concentrations of these seven metabolites were visualized using boxplots via the "ggplot2" function in R software. Subsequently, we selected metabolites that showed a Wilcoxon-Mann-Whitney test p < 0.05 in the targeted cohort and whose orientation was consistent with that in the non-target cohort as the final diagnostic biomarkers. Ultimately, three candidate diagnostic biomarkers were determined as the final diagnostic biomarkers.

[0124] (4.2) Evaluation of the diagnostic model for the final diagnostic standard

[0125] First, we matched the multicenter targeted cohorts according to demographic information to create a targeted training set and a targeted prediction set. Then, using the "glm" function in R software, we constructed multiple logistic regression models based on three final diagnostic biomarkers using the targeted training set. Subsequently, in both the targeted training and targeted prediction sets, we evaluated the diagnostic power of the logistic regression models based on the three final diagnostic biomarkers using receiver operating characteristic (ROC) curves.

[0126] Furthermore, in the non-target cohort data, we also constructed a logistic regression model based on the final three diagnostic biomarkers using the "glm" function in R software, utilizing the discovery set. Subsequently, we evaluated the diagnostic power of the logistic regression model based on the final three diagnostic biomarkers using receiver operating characteristic (ROC) curves in the discovery set (70% of the training set), the internal validation set (30% of the training set), the validation set, and the test set. Finally, based on the diagnostic model constructed from the targeted training set, we fitted the optimal overall curve-off value using the Youden Index in the targeted training set using the "glm" function in R software.

[0127] (4.3) Correlation analysis and subgroup analysis of clinical information and three diagnostic markers

[0128] The workflow for diagnostic biomarker correlation and subgroup analysis is as follows: Figure 1As shown in Figure 1D. We used the "rcorr" function in R software to perform Spearman correlation analysis on the correlation between clinical indicators (grouping, age, BMI, sex, HAMD, and HAMA) and three final diagnostic markers in three non-target cohorts and one targeted cohort. We found a stable correlation between sex and unsaturated undecenoic acid carnitine (Car (11:1)). Therefore, we further performed statistical analysis on the sex subgroups of participants in the non-target training set, non-target validation set, non-target test set, and targeted cohort using the Kruskal-Wallis test. Finally, we also attempted to use the Kruskal-Wallis test to compare the differences between the depressed and normal sex subgroups among participants from different centers.

[0129] (4.4) Optimal subgroup specificity cut-off value analysis of the diagnostic model based on Youden Index

[0130] Based on the "glm" function in R, we tested the sensitivity and 1-specificity of the optimal overall cut-off value across different grouping types: the entire target cohort, the target training set, the target prediction set, gender-specific groups (male and female), and groups with different inclusion centers (Chong Qing, Chang Sha, and Nan Jing). Finally, we fitted the subgroup-specific cut-off value using the Youden Index in different subgroups (gender-specific and group with different inclusion centers) and compared the model's sensitivity and 1-specificity across each subgroup under both the overall and subgroup-specific cut-off values.

[0131] (4.5) Data Statistical Analysis

[0132] Participant demographic and clinical characteristics were analyzed using SPSS. Gender differences were analyzed using the chi-square test, while differences in age, BMI, HAMA score, and HAMD score were analyzed using the Wilcoxon-Mann-Whitney test. All statistical tests in this study were two-tailed.

[0133] (5) Validation of the targeted metabolism differences of seven candidate diagnostic biomarkers

[0134] Through screening in the non-target cohort, we identified seven plasma metabolites—cyclic adenosine monophosphate (cAMP), methylpyridine, unsaturated undecenoic carnitine (Car(11:1)), histidine, prolyl hydroxyproline, tryptophan, and glucuronic acid—as candidate diagnostic biomarkers. Subsequently, we performed absolute quantitative targeted metabolomics validation of these seven metabolites using isotope-labeled standards. Based on the criteria of meeting the Wilcoxon-Mann-Whitney test (P < 0.05) in the targeted data and the consistency of the fold change (FC) direction with that in the non-target cohort, we screened three metabolites: cAMP (Down, P < 0.001), unsaturated undecenoic carnitine (Car(11:1) (Down, P < 0.001), and glucuronic acid (Up, P < 0.001)—as the final diagnostic biomarkers. Figure 7 AC). Among the other four candidate metabolites, there were no significant differences in methylpyridine (P = 0.125), histidine (P = 0.503), and tryptophan (P = 0.078). While there were differences in prolyl hydroxyproline (P = 0.006), the trend of change was opposite to that in the non-target cohort. Figure 7 These experimental data also demonstrate the accuracy and reliability of targeted metabolomics absolute quantitative studies compared to non-targeted metabolomics studies. This technical approach is the first targeted metabolomics study targeting plasma biomarkers for depression in children and adolescents, and it is the first to propose a combined diagnostic biomarker approach using cyclic adenosine monophosphate, unsaturated undecenoic acid carnitine (Car (11:1)), and glucuronic acid, without existing technology for reference. Targeted validation can clarify the objective absolute concentration of metabolites, making the diagnostic formula more reliable and enabling direct clinical application. In addition, proline and hydroxyproline have been reported to change in the plasma of adult patients with depression in some literature, and histidine metabolism, tryptophan metabolism, and adult depression are associated. However, the changes in amino acid metabolites in children and adolescents with depression differ from those in adults, and no difference in these substances between children and adolescents with depression and normal populations has been shown. This is a unique characteristic of children and adolescents with depression, and the results of this approach cannot be inferred from the metabolite changes in adults with depression.

[0135] The literature indicates that in non-adolescent populations (e.g., adults over 30 years of age), there was no significant difference in platelet cAMP levels between MDD patients and healthy individuals (original text: The mean basal cAMP activity was 24.3 ± 13.6 nM cAMP / well in the MDD subjects and 27.1 ± 14.15 nM cAMP / well in the healthy controls (t = -0.87; p = ns)). The literature also states that when platelets in the blood respond to prostaglandin stimulation, cAMP levels in depressed patients are lower than in healthy individuals. The response of platelets in the blood to prostaglandin stimulation (the ability to release cAMP) is an indicator that can diagnose depression and normalcy (differences in adult cAMP levels only occur under prostaglandin administration) and can predict the efficacy of antidepressant treatment. Therefore, in the absence of prostaglandin administration, there is no difference in platelet cAMP levels between MDD patients and healthy individuals, and the cAMP levels in plasma and / or platelets cannot be directly used as a diagnostic marker to distinguish between adult MDD patients and healthy individuals. This technical solution discovered a significant difference in cAMP levels in the plasma of MDD patients and healthy individuals in the specific population of children and adolescents without the application of release stimuli. This difference can be directly used as a diagnostic biomarker, which was unpredictable in existing reports. The existing technical literature mentioned in this paragraph is: Steven D. Targum, A novel peripheral biomarker for depression and antidepressant response, Molecular Psychiatry volume 27, pages 1640–1646 (2022). Our study found a significant decrease in cAMP levels in the plasma of children and adolescents with depression. However, previous adult metabolomics studies did not find changes in cAMP levels in the plasma of depressed individuals. The inventors analyzed that the reason for this phenomenon may be that children and adolescents are in a critical period of rapid growth and development. Growth and development are accompanied by organ and body remodeling, requiring a large amount of cell breakdown and proliferation. Cell breakdown exacerbates the metabolism of purines into uric acid and its excretion, while cell proliferation requires a large amount of purines for gene amplification and transcription. Both of these factors increase the body's purine consumption and weaken the protective effect of cAMP on the nervous system. Ultimately, the additional purine metabolism requirements of growth and development, combined with adverse psychological stimuli, lead to the development of depression, accompanied by a downregulation of purine metabolism.

[0136] In addition, the inventors' prior patents used non-targeted metabolomics to study biomarkers for various mental illnesses. The results showed that cAMP is a member of a biomarker combination used to distinguish between adolescents with schizophrenia and healthy individuals, and that plasma cAMP levels were not correlated with the diagnosis of depression. This indicates that prior art reports have a negative technical implication for this solution, which overcomes the bias of existing technologies and reveals that cAMP is an ideal biomarker for diagnosing depression in children and adolescents. The inventors' prior patent application is: CN118010881A, a plasma biomarker combination and diagnostic system for diagnosing three mental illnesses in children and adolescents.

[0137] In the same prior patent (CN118010881A), the inventors discovered that glucuronic acid is a member of a group of biomarkers used to distinguish between adolescents with bipolar disorder and healthy individuals, and that plasma glucuronic acid levels are not correlated with the diagnosis of depression. This also indicates that prior art reports have created negative technical inspiration for this technical solution. This solution overcomes the bias of prior art and reveals that glucuronic acid is an ideal biomarker for diagnosing depression in adolescents and children.

[0138] Although a prior patent (CN118010881A) describes unsaturated undecenoylcarnitine (Car11:1) as a biomarker for diagnosing depression, in this study, unsaturated undecenoylcarnitine (Car11:1) needs to be combined with 5-aminoacetylacetonate, ethanol, decanoic acid, salicylic acid, pyruvate, etc., as a biomarker combination to achieve the diagnosis of depression. Whether unsaturated undecenoylcarnitine (Car11:1) alone would still have diagnostic efficacy is not technically apparent from the prior patent. In fact, in this technical solution, using unsaturated undecenoylcarnitine (Car11:1) alone as a diagnostic biomarker is not effective in distinguishing between patients with depression and healthy individuals. The diagnostic efficacy of unsaturated undecenoylcarnitine (Car11:1) alone is even less than that of cAMP or glucuronic acid. Unsaturated undecenoylcarnitine (Car11:1) needs to be used in combination with cAMP and glucuronic acid as a diagnostic biomarker to effectively diagnose depression.

[0139] (6) Evaluation of the diagnostic efficacy of three diagnostic markers

[0140] Based on three final biomarkers, we constructed the equation (Logit(P)=a1×M) using a logistic regression model in the targeted training set. cAMP +a2×M Glucuronic +a3×M Car(11:1)+b), specifically: Logit(P) = -0.2102011 × M cAMP +0.0024516×M Glucuronic -0.0087530×M Car(11:1) +2.5230796 (b is the intercept (constant), a) n Let be the coefficient for each diagnostic biomarker; M represent the quantitatively measured content of the metabolite; P is the risk value (test variable) for depression, 0 ≤ P ≤ 1; P is compared with the cut-off value (threshold) to diagnose whether the sample has depression. The equation was used to fit the ROC curves of the targeted training set and the targeted prediction set. It was found that this absolute quantitative equation has good diagnostic power for both the targeted training set (AUC = 0.849) and the targeted prediction set (AUC = 0.858). Figure 8 A). Furthermore, in the targeted training set, the optimal cut-off value fitted using the Youden Index is 0.534, which demonstrates good diagnostic performance in both the targeted training set (1-Specificity = 0.800, Sensitivity = 0.730) and the targeted prediction set (1-Specificity = 0.840, Sensitivity = 0.690). Figure 8 A).

[0141] In addition, the inventors also used cAMP alone as a biomarker, generating ROC curves for the targeted training and prediction sets. The AUC for the targeted training set was 0.776, and the AUC for the targeted prediction set was 0.867. Based on this, cAMP and glucuronic acid were combined as diagnostic biomarkers. A logistic regression model was used to construct the equation, generating ROC curves for the targeted training and prediction sets. The AUC for the targeted training set was 0.838, and the AUC for the targeted prediction set was 0.856. Furthermore, cAMP, glucuronic acid, and unsaturated undecenoic acid carnitine (Car (11:1)) were combined as diagnostic biomarkers. A logistic regression model was constructed to construct the equation, generating ROC curves for the targeted training and prediction sets. The AUC for the targeted training set was 0.849, and the AUC for the targeted prediction set was 0.858.

[0142] In addition, ROC curves were generated for the training and prediction sets of glucuronic acid as a single biomarker, with an AUC of 0.698 for the training set and 0.580 for the prediction set. ROC curves were also generated for unsaturated undecenoylcarnitine (Carb(11:1)) as a single biomarker, with an AUC of 0.669 for the training set and 0.716 for the prediction set.

[0143] The experimental results above show that cAMP has the best diagnostic effect when used alone as a diagnostic biomarker, along with glucuronic acid or unsaturated undecenoic acid carnitine (Car11:1). cAMP achieves AUC values ​​>0.7 or even >0.8 in both the targeted training and prediction sets, meeting the requirements for the diagnostic model. However, glucuronic acid or unsaturated undecenoic acid (Car11:1) struggles to achieve an AUC value of 0.7 in the targeted training set, indicating that their effectiveness as standalone diagnostic biomarkers is not ideal. To further improve diagnostic performance, the inventors constructed equations using a logistic regression model, combining two or three diagnostic biomarkers. The results showed that the combined use of cAMP and glucuronic acid effectively increased the AUC value to above 0.8, while the combined use of all three biomarkers further improved the AUC value, resulting in a more ideal diagnostic model.

[0144] Among them, the ROC curve (Receiver Operating Characteristic curve) is used to evaluate the quality of classification and detection results. It is a very important and common statistical analysis method. It is a coordinate graph composed of the false positive rate (1-specificity) on the horizontal axis and the true positive rate (sensitivity) on the vertical axis. The curve is plotted to show the different results obtained by the test sample under different judgment criteria (thresholds). The area under the curve (AUC) is used to represent accuracy. The higher the AUC value, the higher the accuracy. An AUC value > 0.7 indicates that the diagnostic model can diagnose diseases relatively accurately. The closer the value is to 1, the more ideal the diagnostic model is.

[0145] Furthermore, based on three final diagnostic biomarkers, we constructed an equation using a logistic regression model in the non-target discovery set. Using this equation, we fitted the ROC curves of the non-target discovery set (70% of the non-target training set) (AUC = 0.878), the internal non-target validation set (30% of the non-target training set) (AUC = 0.863), the non-target validation set (AUC = 0.793), and the non-target test set (AUC = 0.833), all achieving good diagnostic efficacy. Figure 8 BC).

[0146] In addition to joint diagnosis targeting three biomarkers, joint diagnosis targeting seven biomarkers was also performed, and equations were constructed using a logistic regression model in the targeted training set:

[0147] Logit(P)=-0.2722×M1+0.002628×M2-0.01476×M3+0.002021×M4+0.0005301×M5+0.000555×M

[0148] 6 + 0.000008763 × M7 - 0.8001. M1-M7 are variables, used sequentially to input the plasma concentrations of cyclic adenosine monophosphate (cAMP), glucuronic acid, unsaturated undecylcarnitine, methylpyridine, histidine, prolyl hydroxyproline, and tryptophan. This absolute quantitative equation was found to have good diagnostic power for both the training set (AUC = 0.8807) and the prediction set (AUC = 0.8375).

[0149] (7) Correlation analysis between three diagnostic markers and clinical information

[0150] Next, based on the three final diagnostic biomarkers, we used Spearman correlation analysis in three non-target cohorts and one targeted cohort to explore the correlation between clinical indicators (grouping, age, BMI, sex, HAMD, and HAMA) and the three final biomarkers. We ultimately found a stable correlation between sex and unsaturated undecylcarnitine (Car(11:1)): In the non-target training set, non-target validation set, non-target test set, and the targeted cohort, among the total population (MDD & HC), depressed patients (MDD), and normal controls (HC), a total of 12 combinations, only the depressed patients in the targeted cohort did not show a significant correlation; the other 11 combinations all showed significant correlations. Figure 9 Additionally, we found that the three metabolites were correlated with grouping information and clinical symptom scale scores (HAMD and HAMA) at the overall population level (MDD & HC), but the correlation between the three metabolites and clinical symptom scale scores (HAMD and HAMA) was not significantly different between isolated patients with depression (MDD) and normal controls (HC). Figure 9 (AD). This suggests that the correlation between the three metabolites and clinical symptom scale scores (HAMD and HAMA) is due to the significant differences in clinical symptoms among different groups.

[0151] (8) Sex subgroup analysis of carnitine (Carb) (11:1)

[0152] Because correlation analysis revealed a stable correlation between carnitine Carb(11:1) and gender, we performed Kruskal-Wallis statistical analysis on different gender subgroups (depressed and normal) in the non-target training set, non-target validation set, non-target test set, and targeted cohort. The results showed that in all four groups (non-target training set, non-target validation set, non-target test set, and targeted cohort), normal males had significantly higher carnitine Carb(11:1) levels than normal females, and depressed males had significantly lower levels than normal males, while depressed females also had significantly lower levels than normal females. Figure 10However, in depressed men, carnitine (Carbine Carbide ... Figure 10 D).

[0153] The difference in carnitine levels between the non-target and target populations and between different genders in depressed patients is unclear. It is uncertain whether this is related to the fact that the target cohort came from three different centers, while the non-target cohort was entirely from the Chong Qing region. Therefore, further clarification is needed regarding the differences in carnitine changes among depressed patients from different sources. We used the Kruskal-Wallis test to statistically compare gender subgroups within the depressed and normal groups from different centers. We found that in the non-target population from Chong Qing, male patients with carnitine Car (Car11:1) were significantly higher than female patients, but no difference was found in the target population. Figure 11 AB). Among the target population of depressed patients derived from Chang Sha, there was no difference in carnitine levels between genders. Figure 11 C). However, in the targeted population of depressed patients derived from Nan Jing, the carnitine Carbide (Carb) levels in men (11:1) were significantly higher than in women (C). Figure 11 D).

[0154] (9) Sex-specific cut-off values ​​and inclusion center subgroups

[0155] Through correlation and subgroup analyses, we found that different genders and population origins may affect the extent of downregulation of the diagnostic biomarker for depression (carnitine (Carbine 11:1)). Therefore, we attempted to find subgroup-specific cut-off values ​​in our absolute quantitative diagnostic model based on gender subgroups and subgroups of the center of origin of the enrolled population. We found that the cut-off values ​​fitted solely to males had better diagnostic performance than the global cut-off values ​​directly applied to males, and the cut-off values ​​fitted solely to samples from Chang Sha had better diagnostic performance than the global cut-off values ​​directly applied. The diagnostic performance of cut-off values ​​fitted to other subgroups was not significantly different from that of the global cut-off values ​​(Table 5).

[0156] Table 5: Cut-off values ​​specific to sex subgroups and inclusion center subgroups

[0157]

[0158] Example 4: Evaluation of the combined diagnostic efficacy of gender and three diagnostic markers

[0159] Based on three final biomarkers and a sex variable, we constructed the equation (Logit(P) = a1 × M) using a logistic regression model on the targeted training set.cAMP +a2×M Glucuronic +a3×M Car(11:1) +a4×Sex+b), specifically: Logit(P)=-0.219851×M cAMP +0.002460×M Glucuronic -0.011070×M Car(11:1) -0.671532×Sex+4.026209 (b is the intercept, a) n denoted as the coefficient for each diagnostic biomarker; M represents the quantitatively measured content of the metabolite, and Sex is set to male = 1, female = 2). The equation was used to fit ROC curves for the targeted training set and the targeted prediction set. The absolute quantitative equation showed good diagnostic power for both the targeted training set (AUC = 0.855) and the targeted prediction set (AUC = 0.852). Furthermore, in the targeted training set, the optimal cut-off value fitted using the Youden Index was 0.490, which demonstrated good diagnostic performance in both the targeted training set (1-Specificity = 0.800, Sensitivity = 0.790) and the targeted prediction set (1-Specificity = 0.810, Sensitivity = 0.770). After adding the sex variable, the diagnostic performance of the subgroup fitted cut-off value and the global cut-off value was not significantly different (Table 6).

[0160] Table 6: Diagnostic model assessment of sex-related factors involving three metabolites

[0161]

[0162]

[0163] For the combined sex diagnosis using seven biomarkers, a logistic regression model was used to construct the equation in the targeted training set: Logit(P)=-0.2756×M1+0.002628×M2-0.001359×M3+0.001737×M4+0.0006020×M5+0.0007109×M6+0.000009975×M7+0.5543×M8-2.336. M1-M7 are variables, used to input the plasma concentrations of cyclic adenosine monophosphate, glucuronic acid, unsaturated undecylcarnitine, methylpyridine, histidine, prolyl hydroxyproline, and tryptophan, respectively; M8 is a variable, used to input sex information when considering sex factors, with a value of 1 for males and 2 for females. The absolute quantitative equation was found to have good diagnostic power for both the target training set (AUC = 0.8806) and the target prediction set (AUC = 0.8376).

[0164] Example 5: Other available diagnostic biomarkers and combinations of diagnostic biomarkers

[0165] In this study, in addition to discovering that cyclic adenosine monophosphate (cAMP), a combination of cAMP and glucuronic acid, a combination of cAMP, glucuronic acid, and unsaturated undecenoic acid carnitine, and a combination of cAMP, cAMP, unsaturated undecenoic acid carnitine, histamine, prolyl hydroxyproline, tryptophan, and glucuronic acid can serve as biomarkers for the diagnosis of depression in adolescents and children, other combinations of these seven substances can also be used for the diagnosis of depression in adolescents and children, achieving good diagnostic results. In this field, a diagnostic model with an AUC value > 0.7 is generally considered to have the potential for further application.

[0166] In this technical solution, a logistic regression model is constructed using datasets obtained from targeted and non-targeted metabolomics. The dataset includes several records (samples) from children and adolescents with depression or normal children and adolescents. Each record includes the name and content information of the corresponding diagnostic biomarkers. If gender factors are to be included in the diagnostic model, each record also includes gender information. Following conventional methods, the logistic regression model based on different biomarkers (combinations) (or based on different biomarkers (combinations) and gender information) is trained using the "glm" function in R software. The diagnostic efficacy of the logistic regression model based on different biomarkers (combinations) is evaluated based on receiver operating characteristic (ROC) curves (evaluation results are detailed in Tables 7 and 8). The key point of this study is the establishment of seven biomarkers (and their combinations) through metabolomics research, as well as the identification of gender influencing factors. These are crucial technical points for establishing a diagnostic model and achieving effective diagnosis. The training, fitting, modeling, and efficacy evaluation processes based on the aforementioned biomarkers and gender information are all conducted using conventional methods.

[0167] More specifically, the logistic regression model is as follows:

[0168] Logit(P)=a1×M1+a2×M2+a3×M3+a4×M4+a5×M5+a6×M6+a7×M7+a8×M8+b;

[0169] Among them, M1-M7 are independent variables, which are used to input the numerical values ​​of the plasma content of cyclic adenosine monophosphate, glucuronic acid, unsaturated undecenoylcarnitine, methylpyridine, histidine, prolyl hydroxyproline and tryptophan, respectively; among the independent variables of M1-M7, at least one independent variable is input with the numerical value of the corresponding metabolite content, and the remaining independent variables are set to 0.

[0170] M8 is the independent variable. When considering gender factors, M8 is used to input gender information, with a value of 1 for males and 2 for females; when not considering gender factors, M8 is set to 0.

[0171] a1-a8 are the coefficients of M1-M8 respectively, and b is a constant; the specific values ​​of a1-a8 and b can be obtained through the above specific process of constructing the logistic regression model.

[0172] 0≤P≤1, the diagnosis is achieved by comparing the P value with the critical value.

[0173] Table 7: AUC values ​​of different diagnostic biomarkers (combinations) in logistic regression models considering or not considering gender factors in targeted metabolomics data.

[0174]

[0175]

[0176]

[0177] Table 8: AUC values ​​of different diagnostic biomarkers (combinations) in logistic regression models considering or not considering gender factors in non-targeted metabolomics data.

[0178]

[0179]

[0180]

[0181]

[0182]

[0183] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. The application of plasma metabolic diagnostic markers in constructing a diagnostic system for childhood and adolescent depression, characterized in that, The plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate (cAMP) and group A markers, or consist of cAMP, group A markers, and group B markers. Group A biomarkers include at least one of unsaturated undecenoylcarnitine, methylpyridine, and prolyl hydroxyproline; Group B biomarkers include at least one of histidine, tryptophan, and glucuronic acid. The structural formula of the unsaturated undecenoylcarnitine is: 。 2. The application of the plasma metabolic diagnostic markers according to claim 1 in constructing a diagnostic system for childhood and adolescent depression, characterized in that, The plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate, glucuronic acid, and unsaturated undecenoylcarnitine.

3. The application of the plasma metabolic diagnostic markers according to claim 1 in constructing a diagnostic system for childhood and adolescent depression, characterized in that, The plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate, methylpyridine, unsaturated undecenoylcarnitine, histidine, prolyl hydroxyproline, tryptophan, and glucuronic acid.

4. A diagnostic system for childhood and adolescent depression, characterized in that, It includes a data acquisition unit and a diagnostic unit; The data acquisition unit includes a content detection subunit, or includes a content detection subunit and a gender information acquisition subunit; the content detection subunit is used to detect the content of plasma metabolic diagnostic markers in plasma; The plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate (cAMP) and group A markers, or consist of cAMP, group A markers, and group B markers. Group A biomarkers include at least one of unsaturated undecenoylcarnitine, methylpyridine, and prolyl hydroxyproline; Group B biomarkers include at least one of histidine, tryptophan, and glucuronic acid. The structural formula of the unsaturated undecenoylcarnitine is: ； The diagnostic unit is used to obtain the test variable value P for the risk of diagnosing depression through a logistic regression model. The logistic regression model is: Logit(P) = a1×M1 + a2×M2 + a3×M3 + a4×M4 + a5×M5 + a6×M6 + a7×M7 + a8×M8 + b; Among them, M1-M7 are independent variables, which are used to input the numerical values ​​of the plasma content of cyclic adenosine monophosphate, glucuronic acid, unsaturated undecenoylcarnitine, methylpyridine, histidine, prolyl hydroxyproline and tryptophan, respectively; among the independent variables of M1-M7, at least one independent variable is input with the numerical value of the corresponding metabolite content, and the remaining independent variables are set to 0. M8 is the independent variable. When considering gender factors, M8 is used to input gender information, with a value of 1 for males and 2 for females; when not considering gender factors, M8 is set to 0. a1-a8 are the coefficients of M1-M8 respectively, and b is a constant; 0≤P≤1, the diagnosis is achieved by comparing the P value with the critical value.

5. The diagnostic system for childhood and adolescent depression according to claim 4, characterized in that, The logistic regression model is constructed by the following method: collecting information on the content of plasma metabolic diagnostic markers in plasma from several samples, or simultaneously collecting information on the content of plasma metabolic diagnostic markers in plasma and gender information from several samples, thereby obtaining a dataset; using the logistic regression algorithm to train the model on the dataset to obtain the logistic regression model; The critical value is obtained by fitting an ROC curve using a logistic regression model in the dataset. The value of the test variable corresponding to the maximum value of the Youden index of the ROC curve is the critical value used for diagnosis.

6. The diagnostic system for childhood and adolescent depression according to claim 4, characterized in that, Plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate, glucuronic acid, and unsaturated undecenoylcarnitine: Without considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula: Logit(P)=-0.2102011×M1+0.0024516×M2-0.0087530×M3+2.5230796, Alternatively, considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula: Logit(P)=-0.219851×M1+0.0024516×M2-0.011070×M3-0.671532M8+4.026209.

7. The diagnostic system for childhood and adolescent depression according to claim 4, characterized in that, Plasma metabolic diagnostic markers consist of cyclic adenosine monophosphate, glucuronic acid, unsaturated undecenoylcarnitine, methylpyridine, histidine, prolyl hydroxyproline, and tryptophan; Without considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula: Logit(P) = -0.2722 × M1 +0.002628×M2-0.01476×M3+0.002021×M4+0.0005301×M5+0.000555×M6+0.000008763×M7-0.8001; Alternatively, considering gender, the test variable P for the risk of diagnosing depression is calculated using the following formula: Logit(P)=-0.2756×M1+0.002628×M2 -0.001359×M3+0.001737×M4+0.0006020×M5+0.0007109×M6+0.000009975×M7+0.5543×M8-2.

336.

8. A diagnostic system for childhood and adolescent depression according to any one of claims 4-7, characterized in that, The content detection subunit includes high-performance liquid chromatography. Mass spectrometry equipment.

Images