Biomarker for identifying multiple omics subtypes of major depression and application of biomarker

Abstract

The invention relates to a biomarker for identifying multiple omics subtypes of major depression and application of the biomarker. The multiple omics subtypes of major depression are respectively a MoS1 subtype, a MoS2 subtype and a MoS3 subtype. The biomarker for identifying the MoS1 subtype comprises the following components: IGFBP-2, Met, classicalMonocytes, B, Spermidine, GFBP-2, GFBP-2, GFBP-2, GFBP-2, GFBP-2, GFBP-2, GFBP-2 The biomarker for identifying the MoS2 subtype comprises Spermidine, SDMA (Space Division Multiple Access), TG.16.1332. 2., Cys, TG.16.1332. 2., TG.16.1332. 2., the biomarker for identifying the subtype MoS3 comprises the following components: Monocytes, CD4, HLADRMonocytes, Activated Treg, and CD36 Treg, and can be used for identifying the subtype MoS3. The invention also discloses a method for identifying the subtype MoS3 by using the biomarker for identifying the subtype MoS3. Good sensitivity and specificity balance is achieved under the optimal threshold value, and good diagnosis efficiency is shown.

Description

Technical Field

[0001] This invention belongs to the field of mental illness diagnosis technology, and particularly relates to the field of major depressive disorder diagnosis technology, specifically to biomarkers for identifying multi-omics subtypes of major depressive disorder and their applications. Background Technology

[0002] Major depressive disorder (MDD) is a prevalent and highly disabling mental disorder worldwide, affecting more than 300 million people and imposing a heavy burden on individuals and society. However, depression is not a single disease entity, but rather a syndrome with significant clinical and biological heterogeneity. Its etiology involves diverse pathophysiological mechanisms and manifests as a broad spectrum of symptoms. This heterogeneity leads to insufficient precision in the current diagnostic framework, thereby affecting the prediction of treatment response and the selection of intervention strategies. Therefore, identifying subtypes of depression with different biological bases is crucial for achieving more accurate diagnosis and individualized treatment.

[0003] Relying on a single or limited number of biomarkers is insufficient to fully reveal the high heterogeneity of depression. Multi-omics technologies, by integrating multi-level biological information, provide a powerful tool for analyzing the heterogeneity of this disease and identifying its potential subtypes. Currently, research using multi-omics data to explore depression subtypes remains very limited. Hagenberg et al. integrated 43 plasma immune markers with transcriptomic data to identify four subgroups, two of which were closely related to immune-related symptom clusters. Joyce et al. found that combining metabolomics and pharmacogenomics data was more effective in predicting antidepressant treatment response than using metabolomics alone. Tang et al. defined three subtypes of major depressive disorder based on neuroimaging and further elucidated their underlying biological characteristics using multi-omics data. Although some studies have preliminarily identified some depression subtypes using multi-omics methods, effectively integrating multidimensional data to promote personalized diagnosis, treatment, and prognostic assessment of depression remains a key challenge.

[0004] Mounting evidence suggests that the development and progression of depression are closely related to inflammatory responses, immune dysfunction, and metabolic disorders. Significant immunometabolic dysregulation is observed in approximately 20%-30% of patients with depression. Metabolic abnormalities and dyslipidemia can exacerbate neuroinflammation through mechanisms such as vascular dysfunction and oxidative stress, thereby worsening depressive and anxiety symptoms. At the inflammatory level, studies have found an association between elevated peripheral blood cytokine levels in patients with depression and their response to antidepressant treatment. These studies suggest that incorporating inflammation, immune, and metabolic regulation into personalized antidepressant treatment strategies holds significant potential.

[0005] The Peripheral Biomarker Multiomics Integration Module (iMORE) study is a prospective observational cohort study established by the applicant's research team to systematically integrate multiomics data, including metabolomics, cytokines, and immunophenotypes. Based on this, the iMORE cohort's multiomics data will be used to conduct subtype analysis of depression to explore its relationship with treatment response and provide a basis for developing new treatment strategies. Summary of the Invention

[0006] The main objective of this invention is to address the above-mentioned problems by providing a biomarker for identifying multi-omics subtypes of major depressive disorder and its application.

[0007] To achieve the above objectives, a first aspect of the present invention provides a biomarker for identifying multiple omics subtypes of major depressive disorder, characterized in that the multiple omics subtypes of major depressive disorder are MoS1 subtype, MoS2 subtype and MoS3 subtype. Biomarkers used to identify MoS1 subtypes include: IGFBP-2, Met, classical Monocytes, B, and Spermidine. Biomarkers used to identify MoS2 subtypes include: Spermidine, SDMA, TG.16.1_32.2, and Cys; Biomarkers used to identify MoS3 subtypes include: Monocytes, CD4, HLADR_Monocytes, Activated_Treg, and CD36_Treg.

[0008] Preferably, there are significant differences in the efficacy of antidepressant treatment among different subtypes. The MoS1 subtype showed the highest efficacy, followed by the MoS3 subtype, while the MoS2 subtype showed the lowest efficacy. The antidepressants used were commonly used medications, such as selective serotonin reuptake inhibitors (SSRIs, such as fluoxetine, paroxetine, sertraline, citalopram, and escitalopram) and serotonin-norepinephrine reuptake inhibitors (SNRIs, such as venlafaxine and duloxetine).

[0009] The full English and Chinese names of each factor are shown in Table 1 below.

[0010] A second aspect of the invention provides the use of reagents for detecting the levels of the said biomarkers in the preparation of products for identifying multi-omics subtypes of major depressive disorder.

[0011] Better place, Diagnostic models for MoS1 subtypes ; Diagnostic models for MoS2 subtypes ; Diagnostic models for MoS3 subtypes .

[0012] The biomarkers for identifying multi-omics subtypes of major depressive disorder and their applications of the present invention achieve a good balance between sensitivity and specificity at the optimal threshold, demonstrating good diagnostic efficacy. Attached Figure Description

[0013] Figure 1 This is a schematic diagram illustrating the results of identifying molecular subtypes through multi-omics data integration. Figure 1 Figure A shows the results of evaluating the optimal number (k=3) for multi-omics clustering; Figure 1 B is the profile analysis result confirming the stability of the subtype, which reached the highest average profile score of 0.62 when k=3; Figure 1 C represents a molecular clustering heatmap based on multi-omics analysis.

[0014] Figure 2 The ROC curves for the MoS1, MoS2, and MoS3 diagnostic models are shown for the training set and the total sample set. All models demonstrate good diagnostic efficacy. The horizontal axis represents 1-specificity (false positive rate), the vertical axis represents sensitivity (true positive rate), and the diagonal line (dashed line) represents the reference line with no discriminative ability (AUC = 0.5).

[0015] Figure 3 This is a graph comparing the clinical outcomes of the three MoS subtypes. Figure 3 A showed that after filling the HAMD-17 score with predicted mean matching, MoS2 had the worst response at both weeks 4 and 8, while MoS1 had the best efficacy. Figure 3 B shows the sensitivity analysis, confirming the robustness of the results; MoS2 still performed the worst, while MoS1 performed the best. Detailed Implementation

[0016] To provide a clearer understanding of the technical content of this invention, the following embodiments are provided in detail. However, it is important to note that these descriptions are merely for further illustrating the features and advantages of this invention, and not for limiting the scope of the claims.

[0017] Unless otherwise specified, the reagents and methods involved in the examples are all commonly used in the art.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] Research subjects This study included 134 patients with myocardial infarction (MDD) and 50 healthy controls. The participants were primarily drawn from the iMORE cohort (clinical trial registration number: NCT04518592). Participants were recruited from the Shanghai Mental Health Center between December 2020 and June 2022.

[0020] Patient inclusion criteria were: adult patients aged 18–65 years with a baseline Hamilton Depression Rating Scale (HAMD-17) score ≥20 and meeting the diagnostic criteria for major depressive disorder (MDD) in the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) (DSM-5).

[0021] Patient exclusion criteria include: having any other Axis I mental disorders, having a high risk of suicide, having a serious physical illness (such as cerebrovascular disease), or being a pregnant or lactating woman.

[0022] This research protocol was approved by the Ethics Committee of Shanghai Mental Health Center, and all participants signed written consent forms after obtaining informed consent.

[0023] Research Methods Clinical data collection and follow-up The main prescription medications for MDD patients are selective serotonin reuptake inhibitors (SSRIs, such as fluoxetine, paroxetine, sertraline, citalopram, and escitalopram) or serotonin-norepinephrine reuptake inhibitors (SNRIs, such as venlafaxine and duloxetine).

[0024] The severity of depressive symptoms in patients with MDD was assessed using the Hamilton Depression Rating Scale-17 (HAMD-17) at baseline (w0), 4 weeks (w4), and 8 weeks (w8). Efficacy was assessed based on the rate of reduction in the total HAMD score after treatment; a reduction of ≥50% was considered effective.

[0025] Mass spectroscopy flow cytometry analysis Mass cytometry was used to analyze 63 types of immune cells. Peripheral blood mononuclear cells were isolated, labeled with isotope barcodes, and mixed with EQ-calibrated microspheres. Cells were then subjected to cisplatin-active staining, Fc receptor blocking, and surface antibody staining. Subsequently, they were fixed, permeabilized, and stained with intracellular antibodies. The stained cells were resuspended in buffer containing Ir-chelating agents and incubated overnight at 2–8°C. Before flow cytometry, the cells were diluted, filtered, and collected at a rate of 200–300 events / second using a Helios mass cytometer. Individual cells. Data were analyzed using Cytobank software, and orthogonal partial least squares discriminant analysis (OPLS-DA) was used to screen for differentially expressed immune cells between groups (projection importance value (VIP) > 1 was the significance threshold).

[0026] Cytokine antibody array detection The levels of 440 cytokines were detected using an antibody array. Plasma samples diluted 1-fold were incubated overnight at 4°C with a chip coated with 440 primary antibodies. After washing, biotin-labeled antibodies were added and incubated for 2 hours. After washing again, Cy3-labeled streptavidin was added and incubated in the dark for 1 hour. Fluorescence signals were acquired using an InnoScan 300 microarray scanner and quantified using Mapix software. Differentially expressed cytokines between MDD and the control group were screened using OPLS-DA.

[0027] Targeted metabolomics detection Plasma samples were analyzed using the MxP® Quant 500 kit according to the instructions. Four replicate quality control pooled plasma samples were included per plate. Approximately 630 endogenous and microbial-derived metabolites were quantitatively detected using an ultra-high performance liquid chromatography-tandem mass spectrometry platform. Metabolites differentially expressed between MDD and control groups were screened using OPLS-DA.

[0028] Multi-omics integration and molecular subtype classification Multi-omics clustering was performed using the "MOVICS" R package. To determine the optimal number of clusters, in addition to calculating the cluster predictive index (CPI) and gap statistic, silhouette scores were calculated based on sample similarity between subtypes. The average silhouette score was used to evaluate the overall quality of the clustering. Ten state-of-the-art multi-omics clustering algorithms were applied (iClusterBayes, SNF, moCluster, PINSPlus, CIMLR, NEMO, IntNMF, COCA, ConsensusClustering, and LRA). For each algorithm, "MOVICS" constructed a patient co-association matrix. , where represents the number of patients, and t corresponds to one of the ten algorithms (2 ≤ t ≤ 10). In this matrix, A non-zero value is only valid if patients A and B are assigned to the same cluster; otherwise, the value is zero. After obtaining results from all ten methods, MOVICS merges these matrices into a single consensus matrix. ,in By integrating results from multiple multi-omics clustering algorithms, the consensus matrix provides a robust metric for measuring pairwise similarity of samples, thereby improving the reliability of the resulting subtypes.

[0029] Statistical analysis methods Based on the homogeneity of variance test results of the data among the three subtype groups, ANOVA or Kruskal-Wallis test was used for inter-group comparisons to screen for features with significant differences. Missing HAMD-17 scores at week 4 and week 8 follow-up were imputed using the Predicted Mean Matching (PMM) method from the R language's MICE package. To verify the robustness of the study results, sensitivity analysis was performed by excluding patients who did not complete the week 4 and week 8 follow-up. To identify the most discriminative features for distinguishing each subtype, a three-category "one-to-many" strategy was used for LASSO regression screening (i.e., constructing three binary classification models: MoS1 vs. MoS2+MoS3). Based on the variables screened by LASSO, multivariate logistic regression (stepwise method) was further used to construct the final clinical prediction model. The model's discriminative power was assessed using receiver operating characteristic (ROC) curves and area under the curve (AUC). The model's calibration was assessed using calibration curves.

[0030] Research Results Based on the iMORE cohort, the inclusion of research subjects and data screening This study included 134 patients with MDD and 50 healthy controls.

[0031] All 50 healthy controls underwent CyTOF testing, metabolomics, and cytokine testing at baseline. The control group was only used for differentially expressed molecules comparison with MDD patients and was not included in subsequent depression subtype analysis.

[0032] Of the 134 MDD patients who underwent CyTOF testing, 99 and 84, respectively, completed baseline metabolomics and cytokine testing. One patient was excluded due to testing failure, leaving a total of 83 MDD patients for subsequent subtype analysis of depression.

[0033] For CyTOF, cytokine, and metabolomics data, this invention used the OPLS-DA method to screen for differentially expressed molecules between the MDD group and the control group based on a VIP value greater than 1. This identified 27 immune cell types, 131 cytokines, and 158 differentially expressed metabolites.

[0034] Finally, 83 MDD patients and 316 differentially expressed molecules were included for subsequent subtype analysis.

[0035] Three multi-omics subtypes of depression (MoSs) were discovered. Multi-omics subtype analysis of depression was performed using the R software package "MOVICS".

[0036] like Figure 1As shown in Figure A, based on the CPI and Gap statistics, 3 was selected as the cluster size to characterize molecular differences. Profilometry was further used to evaluate the quality and robustness of these three clusters. The results show that, as... Figure 1 As shown in B, when k=3, the average profile score is the highest at 0.62. The profile width of MoS1 is 0.78, MoS2 is 0.74, and MoS3 is 0.47.

[0037] Therefore, patients were assigned to three MoS subtypes, specifically as follows: Figure 1 As shown in C. The MoS subtypes are generated objectively based on the inherent similarity of patients' baseline multi-omics data, integrating ten multi-omics clustering algorithms using the "MOVICS" R package. A patient is assigned to a subtype because their overall molecular profile is most similar to other patients within that subtype in the integrated consensus similarity metric, and significantly different from patients in other subtypes. After obtaining stable clustering results, the core molecular features that distinguish each subtype from others are summarized by comparing biomarkers that show significant differences between subtypes.

[0038] MoS1 is characterized by the synergistic high expression of growth factors (e.g., bFGF, HGF, TGF-α) and Th2-type cytokines (e.g., IL-5, IL-13).

[0039] The defining characteristic of MoS2 is a significant increase in triglycerides as its core metabolic feature.

[0040] MoS3 is characterized by a unique immune cell pattern: elevated monocyte levels accompanied by decreased levels of T cells and regulatory T cells.

[0041] Lasso regression analysis was used to select predictors for each group. After using ANOVA or Kruskal-Wallis test, significant statistical differences were found in 120 factors among the three groups, as shown in Table 2 below.

[0042] To identify key predictor variables for each subtype, this invention employs LASSO regression to screen variables for each of the three subtypes. The optimal regularization parameter λ (lambda.min) for each model is determined through 10-fold cross-validation. Balancing model goodness of fit with parsimony, a sparse set of variables with non-zero coefficients is ultimately obtained.

[0043] Specifically: Ten variables were screened out for the MoS1 subtype at λ=0.011, including IGFBP-2, Met, Spermidine, ICAM-2, 2B4, Granulysin, classical Monocytes, B, b-NGF and E-Selectin. The MoS2 isotype retains 6 variables at λ=0.021: Spermidine, SDMA, TG.16.1_32.2, TG.16.1_34.1, Met, and Cys. Five variables were selected for the MoS3 subtype at λ=0.010: Monocytes, CD4, HLADR_Monocytes, Activated_Treg, and CD36_Treg.

[0044] Construction of a multivariate logistic regression diagnostic model This invention constructs discriminant models for three subtypes (MoS1, MoS2, MoS3) using multivariate logistic regression, all of which demonstrate excellent predictive performance. The 83 included MDD samples were randomly divided into a training set (n=59) and a test set (n=24) in a 7:3 ratio for internal validation.

[0045] MoS1 ultimately incorporates five factors: IGFBP-2, Met, classical_Monocytes, B, and Spermidine; MoS2 incorporates four factors: Spermidine, SDMA, TG.16.1_32.2., and Cys; MoS3 incorporates five factors: Monocytes, CD4, HLADR_Monocytes, Activated_Treg, and CD36_Treg.

[0046] Based on the selected biomarkers and their regression coefficients, the final logistic regression model equation is constructed as follows: MoS1 diagnostic model: ; MoS2 diagnostic model: ; MoS3 diagnostic model: .

[0047] The model's discriminative performance is evaluated by calculating the area under the AUC curve. This metric measures the model's ability to distinguish between different classes of samples, with a value ranging from 0 to 1; a higher value indicates stronger discriminative ability. Figure 2 As shown, the discriminative performance of the three molecular subtype (MoS) prediction models is stable across different datasets. In the training set, the AUCs of the MoS1, MoS2, and MoS3 models are 0.993, 0.909, and 0.991, respectively. In the test set, the AUCs are 1.000, 0.905, and 0.992, respectively. Across all samples, the AUCs are 0.996, 0.902, and 0.988, respectively. Each model achieves a good balance between sensitivity and specificity at the optimal threshold, demonstrating good diagnostic efficacy.

[0048] Model calibration performance evaluation The model's calibration performance was evaluated using the Hosmer-Lemeshow test (HL test), which assesses the accuracy of the predicted probability by comparing the difference between the predicted event occurrence rate and the actual observed occurrence rate. A p-value greater than 0.05 indicates that there is no significant difference between the predicted probability and the actual probability, and the model is well calibrated.

[0049] The HL test results for the three subgroup models are as follows: In the training set, the p-values ​​for the HL test of the MoS1, MoS2, and MoS3 models are 0.999, 0.741, and 0.999, respectively. In the test set, the p-values ​​are 0.997, 0.836, and 0.142, respectively. In the entire sample, the p-values ​​are 0.999, 0.910, and 0.982, respectively, all significantly greater than the 0.05 significance level, indicating that there is no significant difference between their predicted probabilities and actual event rates.

[0050] The above results indicate that there is no statistically significant difference between the predicted risk probabilities of each model and the observed actual event occurrence rates, and the prediction results are accurate and reliable.

[0051] Based on the aforementioned high AUC value, it can be concluded that the model constructed in this invention not only has strong discrimination ability but also has good calibrability in probability prediction, resulting in high overall quality.

[0052] Differences in treatment response among patients with different MoS subtypes Of the 83 patients with severe depression, 61 completed the 4-week follow-up and 47 completed the 8-week follow-up.

[0053] There were no significant differences in the distribution of SNRIs or SSRIs among the three MoS subtypes in week 4 (p=0.57) and week 8 (p=0.711), indicating that drug use was balanced among the subtypes.

[0054] PMM was used to impute missing HAMD-17 scores at weeks 4 and 8. The MoS1 subtype included 40 patients, with 31 responders (77.50%) and 9 non-responders (22.50%) at week 8; the MoS2 subtype included 18 patients, with 7 responders (38.90%) and 11 non-responders (61.10%); and the MoS3 subtype included 25 patients, with 16 responders (64.00%) and 9 non-responders (36.00%).

[0055] like Figure 3 As shown in Figure A, MoS2 showed the worst response at week 4, while MoS1 showed the best response, with a statistically significant difference (p=0.008). By week 8, MoS2 still showed poor efficacy, while MoS1 and MoS3 showed better efficacy (χ²=8.03, p=0.020).

[0056] like Figure 3 As shown in B, a sensitivity analysis was performed to verify the robustness of the results, excluding patients who did not complete the follow-up. The results showed that the conclusions were consistent with the above: MoS2 still had the worst efficacy at week 8, while MoS1 and MoS3 had better responses (week 4: p=0.078; week 8: p=0.048).

[0057] Therefore, this invention aims to achieve a dual purpose by classifying MDD patients into MoS subtypes using a multi-omics-driven approach: at the application level, to predict the clinical outcomes of MDD patients with conventional antidepressant treatment; and at the mechanistic level, to reveal and explain the biological heterogeneity of depression.

[0058] Specifically, the core biological characteristics of the three subtypes determine their potential therapeutic significance: MoS1 is characterized by growth factors and Th2 cytokines that promote repair, and its immune regulation and repair tendencies may make it more sensitive to conventional antidepressants, predicting a good treatment outcome; MoS2 is characterized by significant increases in triglycerides, and this systemic metabolic dysregulation is likely to interfere with drug response, thus explaining why this subtype is not effective with standard treatment; while MoS3 exhibits a unique immune pattern of increased monocytes and T cell depletion, suggesting that its pathogenesis may be independent of the traditional monoamine pathway, pointing the way for exploring combined treatment strategies such as immune regulation.

[0059] Therefore, the classification scheme provided by this invention not only transforms a heterogeneous patient population into subtypes with clearly defined biological characteristics, but also transforms the abstract concept of "heterogeneity" into concrete predictive information directly related to clinical outcomes. Specifically, according to the molecular subtyping of major depressive disorder provided by this invention, the efficacy of antidepressant treatment varies significantly among different subtypes. The MoS1 subtype shows the highest efficacy, followed by the MoS3 subtype, while the MoS2 subtype shows the lowest efficacy.

[0060] In this specification, the invention has been described with reference to specific embodiments thereof. However, it will be apparent that various modifications and variations can be made without departing from the spirit and scope of the invention. Therefore, this specification should be considered illustrative rather than restrictive.

Claims

1. A biomarker for identifying multi-omics subtypes of major depressive disorder, characterized in that, The aforementioned major depressive disorder subtypes are MoS1, MoS2, and MoS3. Biomarkers used to identify MoS1 subtypes include: IGFBP-2, Met, classical Monocytes, B, and Spermidine. Biomarkers used to identify MoS2 subtypes include: Spermidine, SDMA, TG.16.1_32.2, and Cys; Biomarkers used to identify MoS3 subtypes include: Monocytes, CD4, HLADR_Monocytes, Activated_Treg, and CD36_Treg.

2. The biomarker according to claim 1, characterized in that, The efficacy of antidepressant treatment varies significantly among different subtypes: the MoS1 subtype has the highest efficacy, followed by the MoS3 subtype, while the MoS2 subtype has the lowest efficacy.

3. The application of the reagent for detecting the level of the biomarker described in claim 1 in the preparation of products for identifying multi-omics subtypes of major depressive disorder.

4. The application according to claim 3, characterized in that, Diagnostic models for MoS1 subtypes ; Diagnostic models for MoS2 subtypes ; Diagnostic models for MoS3 subtypes .

Images