Potential biomarkers differentiating major depression from bipolar disorder depressive episodes by Olink Proteomics Authors: Tingting Xie, Yu Su, Xia Feng, Fangming Xu Published: 2026 ABSTRACT Study designA prospective study.BackgroundIn clinical practice, it is challenging to tell major depressive disorder (MD-D) apart from bipolar disorder depressive episodes (BD-D). This study focuses on using inflammatory proteins as biomarkers to differentiate MD-D from BD-D.Materials and methodsAdmission plasma was collected from 30 healthy participants, 30 BD-D patients, and 30 MD-D patients. Ten samples from each group were randomly chosen for analysis using Olink Proteomics, and all samples were confirmed with ELISA. Then, we established a nomogram prediction model to assess the diagnostic potential biomarkers and their cut-off points.ResultsThe study revealed that patients with MD-D exhibited notably elevated levels of IL6, MCP-3, TGF-α, TNFRSF9, and IL17-A compared to healthy control, with their cut-off values being 2.705 pg/ml, 374.5 pg/ml, 51.5 pg/ml, 328.5 pg/ml, and 2.905 pg/ml, respectively. Significantly higher levels of IL17-A and IFN-γ were found in MD-D patients compared to BD-D individuals and their cut-off values were 3.645 pg/ml, and 40.4 pg/ml, respectively. Subsequently, we created a nomogram prediction model that exhibited excellent consistency in the correction curve and proved to be clinically practical based on decision curve analysis.ConclusionsOur research indicated plasma inflammatory proteins associated with MD-D and could serve as potential biomarkers for it. Moreover, we discovered their cut-off values. A prediction model was created by us to accurately predict MD-D from BD-D patients, giving clinicians a novel approach to quickly distinguish MD-D and implement early targeted treatments. Larger samples are also necessary to further verify our results. Introduction Mood disorders are leading contributors to disability, disease, and mortality around the world [1]. Mood disorders, including bipolar disorder (BD) and major depressive disorder (MD-D), are chronic and recurrent, influencing roughly 2–3% and up to 10% of the population, respectively [2, 3]. BD is identified by mood and energy fluctuations, including mania/hypomania and depression, whereas MD-D involves only depressive and dysthymic episodes [1–3]. Approximately 60% of individuals with BD experience symptoms of major depressive episode prior to their initial hypo/manic episode [4, 5]. Consequently, BD is frequently diagnosed late and is often mistaken for MD-D in around 40% of cases [6]. BD patients who are misdiagnosed might take antidepressants, which can raise the risk of mania and result in delay in receiving the correct diagnosis [7, 8]. Thus, it is necessary to differentiate between MD-D and BD for suitable intervention and the initial treatment phases, which are more effective and safer. Recent studies have increasingly shown that immune dysregulations are involved in mood disorders [9–12], indicating that the inflammatory response system’s activation should be seen as a major pathological component [13]. Research has shown that tumor necrosis factor (TNF)-α and interleukin (IL)-6 levels are linked to MD-D [14]. Other studies have found that IL-6, TNF-α, and interferon (IFN)-γ/IL-4 were associated with BD [15]. However, few studies have compared pro-inflammatory cytokine levels in MD-D patients, BD patients, and control subjects. Previous study revealed BD patients with higher amounts of soluble TNF-αR1, soluble IL-6R, and CRP than MD-D patients [16]. Conversely, other research showed higher levels of TNF-α and soluble TNF-αR1 in MD-D but higher levels of IL-6 in BD [17]. Advancements in high-throughput biological technologies, particularly in proteomics, enhance our understanding of these complex disorders by more accurately reflecting dynamic pathophysiological processes [18]. Recent studies using proteomics have identified serum and plasma proteins associated with mood disorders, suggesting a possible future application in clinical treatment for BD depressive episodes (BD-D) or MD-D [19–21]. According to our knowledge, there are few investigations have compared MD-D participants with BD-D patients using Olink Proteomics, a tool for both quantitative and qualitative assessment. In this study, we initially used Olink Proteomics to investigate the differential inflammatory proteins between MD-D and BD-D patients, then we establish a nomogram prediction model to identify the biomarkers of MD-D to distinguish from BD-D. Materials and methods Ethics approval and consent to participate The Institutional Review Board of our hospital (20190404) granted approval for this study before data collection. All patients who participated signed informed consent forms. Subjects This study was performed at the our hospital from April 2023 to April 2024. Plasma samples were gathered from thirty healthy individuals, thirty patients experiencing BD-D, and thirty MD-D subjects, categorized into the healthy control group(HC), BD-D group, and MD-D group, respectively. We randomly selected ten samples from each group to analyze by Olink Proteomics, which were verified by ELISA for all samples. The included criteria for BD-D or MD-D patients: (1) being between 18 and 60 years of age; (2) having a BD-D or MD-D diagnosis according to ICD-10 criteria; (3) patients who have not undergone treatment for at least two weeks. The excluded criteria: (1) patients with other mental diseases; (2) those with inflammatory diseases or metabolic syndromes that could affect inflammatory markers; (3) incomplete data. Healthy subjects met the following included criteria: (1) 18–60 years old; (2) without mental diseases; (3) without diseases that could affect inflammatory markers; (4) completed data. Plasma sample collection From all participants, we drew 5 mL of peripheral venous blood, which was then placed in EDTA tubes. Plasma was isolated by centrifugation at 3000 g for 15 min and kept at -80 °C for subsequent analysis. Analysis of inflammation-related proteins Protein quantification was performed using the Olink® target 92 inflammation panel, as per the manufacturer’s instructions (Olink Proteomics AB, Uppsala, Sweden). Antibody probe pairs labeled with oligonucleotides are allowed to bind to the target proteins in the samples. The binding of two nearby antibodies allows DNA oligonucleotides to hybridize and extend, resulting in the formation of a PCR reporter sequence. Normalized Protein Expression (NPX) is used to indicate protein levels and is measured in arbitrary units on the Log2 scale. The R package ‘OlinkAnalyze’ was employed to detect differentially expressed proteins (DEPs) between groups. Principal component analysis (PCA) was carried out using the ‘princomp’ function to highlight significant data variability. The R package ggplot2 was utilized to create heat maps and volcano plots, alongside performing Gene Ontology (GO) and KEGG enrichment analyses. Additionally, the correlation in expression between the two proteins was examined using Spearman’s correlation. Cytoscape (version 3.9.1) facilitated the construction and visualization of the DEP’s PPI network. ELISA validation According to the manufacturer’s protocol, ELISA kits from Jianglai Biotechnology Co., Ltd. (Shanghai, China) were used to verify peripheral venous blood samples from all participants through ELISA analysis. Statistical analysis We used SPSS (version 25.0, SPSS Inc., Chicago, IL) and R software, with a P-value cut-off of 0.05. Data for continuous variables that met normality assumptions were expressed as mean ± SD (standard deviation) and subjected to the t-test; if they did not, the Mann-Whitney U test was utilized. Chi-square test was performed on count data, and ROC curve analysis helped identify cut-off values for continuous variables. Diagnostic ability was measured by the area under the ROC curve (AUC), which spans from 0 to 100%. During ROC curve analysis, cut-off values for continuous variables are determined by the maximum value of the Youden index, which is sensitivity+specificity-1. Ultimately, the nomogram prediction model for MD-D was developed using R software (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria). R packages ‘survival’, ‘rms’, and ‘hdnom’ were mainly used for model development and validation. Evaluation of the nomogram model’s discrimination was performed through the consistency index (C-index) and ROC analysis. Additionally, calibration curves were plotted to assess the prediction model using the Hosmer-Lemeshow goodness-of-fit test. Results Characteristics of healthy subjects and BD patients We eventually collected peripheral venous admission plasma from all participants. Table 1 presents the characteristics of Olink Proteomics from healthy individuals, BD-D patients, and MD-D patients, while Table 2 shows the characteristics of 30 healthy individuals, 30 BD-D patients, and 30 MD-D patients. No significant differences was found in age, gender, or body mass index. (Tables 1 and 2). Table 1Characteristics of human samples by Olink Proteomics in the three groupsCharacteristicHC(n = 10)MD-D(n = 10)BD-D(n = 10) p Age, years41.5(38.0–46.0)42.0(37.0–43.0)44.5(36.0–45.0)0.980 Gender 1.000male555female555 Body mass index (kg/m2) 22.9(22.4–25.5)24.3(22.4–26.1)24.0(22.9–25.3)0.791*HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Table 1Characteristics of human samples by Olink Proteomics in the three groupsCharacteristicHC(n = 10)MD-D(n = 10)BD-D(n = 10) p Age, years41.5(38.0–46.0)42.0(37.0–43.0)44.5(36.0–45.0)0.980 Gender 1.000male555female555 Body mass index (kg/m2) 22.9(22.4–25.5)24.3(22.4–26.1)24.0(22.9–25.3)0.791*HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Characteristics of human samples by Olink Proteomics in the three groups *HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Table 2Characteristics of human samples by ELISA in the three groupsCharacteristicHC(n = 30)MD-D(n = 30)BD-D(n = 30) p Age, years41.0(37.0–44.0)42.0(39.0–43.0)41.0(38.0–45.0)0.904 Gender 0.875male151416female151614 Body mass index (kg/m2) 24.0(22.4–25.7)23.6(22.0-26.3)23.8(22.7–25.3)0.833*HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Table 2Characteristics of human samples by ELISA in the three groupsCharacteristicHC(n = 30)MD-D(n = 30)BD-D(n = 30) p Age, years41.0(37.0–44.0)42.0(39.0–43.0)41.0(38.0–45.0)0.904 Gender 0.875male151416female151614 Body mass index (kg/m2) 24.0(22.4–25.7)23.6(22.0-26.3)23.8(22.7–25.3)0.833*HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Characteristics of human samples by ELISA in the three groups *HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Figure 1a illustrated the quantity of proteins that were either up-regulated or down-regulated across pairwise comparisons. Figure 1b revealed some repetition among the three groups concerning inflammation-related proteins. PCA was employed to examine the similarity between the three groups. The first component accounted for 13.16% of the variance, while the second component accounted for 10.06%. Fig. 1Comparison of the inflammation-related proteins among major depressive disorder (MD-D), bipolar disorder depressive episodes (BD-D) and health control (HC). a. Histogram shows the number of differential proteins in three groups; b. Principal component analysis (PCA) showed three dimensions among 3 groups differentiated by color. Each point represents a single patient, with patients of similar protein expression profiles positioned next to each other Comparison of the inflammation-related proteins among major depressive disorder (MD-D), bipolar disorder depressive episodes (BD-D) and health control (HC). a. Histogram shows the number of differential proteins in three groups; b. Principal component analysis (PCA) showed three dimensions among 3 groups differentiated by color. Each point represents a single patient, with patients of similar protein expression profiles positioned next to each other Table 3Protein expression analysis among three groupsProteinGroupsP valueHC(n = 10)MD-D(n = 10)BD-D(n = 10)MD-D vs. HCMD-D vs. BD-D4E-BP17.277.297.670.950.36ADA4.954.925.180.850.21ARTN-0.68-0.80-0.670.290.24AXIN15.675.275.590.400.62Beta-NGF1.421.501.550.070.20CASP-81.341.401.450.780.87CCL116.606.986.880.090.63CCL199.789.9410.310.430.29CCL207.758.907.920.170.25CCL2311.4611.5511.660.620.45CCL255.515.375.440.560.80CCL281.241.411.140.460.21CCL35.415.605.470.580.69CCL46.296.206.280.740.76CD2445.024.954.930.740.91CD4010.7710.7510.600.950.59CD54.884.924.990.700.64CD64.564.594.660.790.72CD8A9.619.789.930.550.48CDCP12.432.812.800.170.97CSF-19.089.149.200.610.62CST56.866.486.880.010.06CX3CL13.093.023.120.710.51CXCL18.828.148.330.120.76CXCL107.597.937.780.350.68CXCL117.327.527.150.500.27CXCL510.879.7610.090.080.65CXCL67.737.517.680.490.70CXCL96.267.086.380.080.10DNER8.538.568.510.860.76EN-RAGE1.671.852.350.540.38FGF-198.238.478.720.660.56FGF-215.526.226.700.310.50FGF-231.051.221.330.520.67FGF-51.141.181.120.730.62Flt3L8.607.908.130.000.29GDNF1.571.511.800.560.13HGF7.527.537.680.980.50IFN-gamma7.277.656.750.500.05IL-1 alpha-0.72-0.64-0.600.740.86IL102.102.212.190.560.94IL-10RA0.150.460.500.140.88IL-10RB5.325.385.560.620.25IL-12B5.735.855.860.590.96IL13-0.83-0.68-0.750.220.62IL-15RA0.560.660.740.500.54IL-17 A0.210.580.170.050.03IL-17 C1.962.851.950.150.14IL188.618.758.800.480.78IL-18R16.086.026.450.810.11IL2-0.07-0.30-0.220.130.64IL-200.340.940.460.240.34IL-20RA-0.110.130.160.370.87IL-22 RA10.020.130.340.680.41IL-240.890.851.130.800.07IL-2RB-0.400.43-0.310.150.20IL330.530.450.480.530.81IL4-1.05-0.79-0.720.310.74IL5-0.25-0.140.120.670.40IL61.562.172.140.010.92IL70.780.580.780.350.38IL84.264.504.640.250.60LAP TGF-beta-15.655.385.560.100.37LIF-0.21-0.22-0.270.950.63LIF-R2.552.402.510.220.29MCP-110.5310.8910.940.060.86MCP-28.388.457.780.800.06MCP-30.340.890.780.020.75MCP-413.8913.8613.880.930.98MMP-112.6212.6012.800.960.69MMP-107.437.357.620.700.25NRTN-0.04-0.240.160.440.19NT-32.932.883.000.750.44OPG9.759.459.390.050.65OSM3.193.313.740.710.36PD-L14.784.554.640.280.57SCF9.078.888.790.220.59SIRT24.033.644.030.430.54SLAMF10.890.941.100.630.32ST1A12.602.072.410.070.43STAMBP4.914.765.000.700.65TGF-alpha2.512.772.720.010.60TNF2.912.902.820.950.68TNFB4.223.983.830.320.49TNFRSF94.805.105.100.031.00TNFSF143.713.463.850.340.35TRAIL8.268.238.320.790.45TRANCE4.334.454.580.590.61TSLP-0.50-0.84-0.830.220.95TWEAK9.388.919.070.000.27uPA10.299.8910.090.010.14VEGFA11.0010.9511.150.700.27*HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Table 3Protein expression analysis among three groupsProteinGroupsP valueHC(n = 10)MD-D(n = 10)BD-D(n = 10)MD-D vs. HCMD-D vs. BD-D4E-BP17.277.297.670.950.36ADA4.954.925.180.850.21ARTN-0.68-0.80-0.670.290.24AXIN15.675.275.590.400.62Beta-NGF1.421.501.550.070.20CASP-81.341.401.450.780.87CCL116.606.986.880.090.63CCL199.789.9410.310.430.29CCL207.758.907.920.170.25CCL2311.4611.5511.660.620.45CCL255.515.375.440.560.80CCL281.241.411.140.460.21CCL35.415.605.470.580.69CCL46.296.206.280.740.76CD2445.024.954.930.740.91CD4010.7710.7510.600.950.59CD54.884.924.990.700.64CD64.564.594.660.790.72CD8A9.619.789.930.550.48CDCP12.432.812.800.170.97CSF-19.089.149.200.610.62CST56.866.486.880.010.06CX3CL13.093.023.120.710.51CXCL18.828.148.330.120.76CXCL107.597.937.780.350.68CXCL117.327.527.150.500.27CXCL510.879.7610.090.080.65CXCL67.737.517.680.490.70CXCL96.267.086.380.080.10DNER8.538.568.510.860.76EN-RAGE1.671.852.350.540.38FGF-198.238.478.720.660.56FGF-215.526.226.700.310.50FGF-231.051.221.330.520.67FGF-51.141.181.120.730.62Flt3L8.607.908.130.000.29GDNF1.571.511.800.560.13HGF7.527.537.680.980.50IFN-gamma7.277.656.750.500.05IL-1 alpha-0.72-0.64-0.600.740.86IL102.102.212.190.560.94IL-10RA0.150.460.500.140.88IL-10RB5.325.385.560.620.25IL-12B5.735.855.860.590.96IL13-0.83-0.68-0.750.220.62IL-15RA0.560.660.740.500.54IL-17 A0.210.580.170.050.03IL-17 C1.962.851.950.150.14IL188.618.758.800.480.78IL-18R16.086.026.450.810.11IL2-0.07-0.30-0.220.130.64IL-200.340.940.460.240.34IL-20RA-0.110.130.160.370.87IL-22 RA10.020.130.340.680.41IL-240.890.851.130.800.07IL-2RB-0.400.43-0.310.150.20IL330.530.450.480.530.81IL4-1.05-0.79-0.720.310.74IL5-0.25-0.140.120.670.40IL61.562.172.140.010.92IL70.780.580.780.350.38IL84.264.504.640.250.60LAP TGF-beta-15.655.385.560.100.37LIF-0.21-0.22-0.270.950.63LIF-R2.552.402.510.220.29MCP-110.5310.8910.940.060.86MCP-28.388.457.780.800.06MCP-30.340.890.780.020.75MCP-413.8913.8613.880.930.98MMP-112.6212.6012.800.960.69MMP-107.437.357.620.700.25NRTN-0.04-0.240.160.440.19NT-32.932.883.000.750.44OPG9.759.459.390.050.65OSM3.193.313.740.710.36PD-L14.784.554.640.280.57SCF9.078.888.790.220.59SIRT24.033.644.030.430.54SLAMF10.890.941.100.630.32ST1A12.602.072.410.070.43STAMBP4.914.765.000.700.65TGF-alpha2.512.772.720.010.60TNF2.912.902.820.950.68TNFB4.223.983.830.320.49TNFRSF94.805.105.100.031.00TNFSF143.713.463.850.340.35TRAIL8.268.238.320.790.45TRANCE4.334.454.580.590.61TSLP-0.50-0.84-0.830.220.95TWEAK9.388.919.070.000.27uPA10.299.8910.090.010.14VEGFA11.0010.9511.150.700.27*HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes Protein expression analysis among three groups *HC=health control; MD-D=major depressive disorder; BD-D=bipolar disorder depressive episodes MD-D vs. HC As shown in Table 3, Olink analysis identified 10 DEPs in the MD-D group compared to the HC group (Fig. 1a), which were illustrated in a volcano plot in Fig. 2a. The top 5 increased DEPs (IL6, MCP-3, TGF-α, IL17-A, TNFRSF9) were significantly more elevated in the MD-D than in the HC (Fig. 2b), which were validated by the ELISA (all p < 0.05, Fig. 2c). Moreover, we performed ROC curve analysis, revealing the AUC of IL6 [p<0.0001, AUC area = 0.991, 95%CI(0.976–1.000)], MCP-3 [p<0.0001, AUC area = 0.946, 95%CI(0.893–0.999)], TGF-α [p<0.0001, AUC area = 0.997, 95%CI(0.989–1.000)], TNFRSF9 [p<0.0001, AUC area = 0.984, 95%CI(0.962–1.000)] and IL17-A [p<0.0001, AUC area = 1.000, 95%CI(1.000–1.000)] (Fig. 2d & Table 4). Based on the Youden index, their cut-off values were 2.705 pg/ml, 374.5 pg/ml, 51.5 pg/ml, 328.5 pg/ml, and 2.905 pg/ml, respectively (Table 4). Furthermore, the PPI network indicated that TGF-α and IL6 could be significant in MD-D (Fig. 2e), potentially serving as biomarkers for MD-D. Fig. 2Comparison of the inflammation-related proteins between major depressive disorder (MD-D) and health control (HC). a. Volcano plots showed significant differences of cytokines between MD-D and HC. Proteins highly expressed in MD-D were labeled in red. Proteins highly expressed in HC were labeled in blue; b. Box plot representing differentially expressed proteins between MD-D and HC; c. The expression levels of IL6, MCP-3, TGF-α, TNFRSF9, and IL17-A between MD-D and HC were shown in box plot by ELISA; d. Receiver operating characteristic curve analysis of IL6, MCP-3, TGF-α, TNFRSF9, and IL17-A between MD-D and HC; e. Correlation in protein abundance PPI using String. (* <0.05;**<0.01; ****<0.0001) Comparison of the inflammation-related proteins between major depressive disorder (MD-D) and health control (HC). a. Volcano plots showed significant differences of cytokines between MD-D and HC. Proteins highly expressed in MD-D were labeled in red. Proteins highly expressed in HC were labeled in blue; b. Box plot representing differentially expressed proteins between MD-D and HC; c. The expression levels of IL6, MCP-3, TGF-α, TNFRSF9, and IL17-A between MD-D and HC were shown in box plot by ELISA; d. Receiver operating characteristic curve analysis of IL6, MCP-3, TGF-α, TNFRSF9, and IL17-A between MD-D and HC; e. Correlation in protein abundance PPI using String. (* <0.05;**<0.01; ****<0.0001) Table 4ROC curve analysis and cut-off value between paired groupsVariablesAreaP-value95%CICut-off valueLower limitUpper limit MD-D vs. HC IL60.991<0.00010.9761.0002.705 pg/mlMCP-30.946<0.00010.8930.999374.5 pg/mlTGF-α0.997<0.00010.9891.00051.5 pg/mlTNFRSF90.984<0.00010.9621.000328.5 pg/mlIL17-A1.000<0.00011.0001.0002.905 pg/ml MD-D vs. BD-D IL17-A0.846<0.00010.7470.9443.645 pg/mlIFN-γ0.897<0.00010.8190.97440.4 pg/ml Table 4ROC curve analysis and cut-off value between paired groupsVariablesAreaP-value95%CICut-off valueLower limitUpper limit MD-D vs. HC IL60.991<0.00010.9761.0002.705 pg/mlMCP-30.946<0.00010.8930.999374.5 pg/mlTGF-α0.997<0.00010.9891.00051.5 pg/mlTNFRSF90.984<0.00010.9621.000328.5 pg/mlIL17-A1.000<0.00011.0001.0002.905 pg/ml MD-D vs. BD-D IL17-A0.846<0.00010.7470.9443.645 pg/mlIFN-γ0.897<0.00010.8190.97440.4 pg/ml ROC curve analysis and cut-off value between paired groups GO and KEGG enrichment analyses were used to explore the potential functions of the up-regulated DEPs in the MD-D group. There was an enrichment in positive regulation of inflammatory response, cell-cell signaling, immune response, and cytokine activity (Fig. 3a-b). Additionally, the KEGG enrichment analysis indicated enrichment in cytokine-cytokine receptor interaction, IL-17 signaling pathway, PI3K-Akt signaling pathway, and chemokine signaling pathway (Fig. 3c-d). Fig. 3Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) of the significant inflammation proteins between major depressive disorder (MD-D) and health control (HC). a & b. GO terms of the signiffcant inffammation proteins between MD-D and HC; c & d. KEGG enrichment analysis of the significant inflammation proteins between MD-D and HC Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) of the significant inflammation proteins between major depressive disorder (MD-D) and health control (HC). a & b. GO terms of the signiffcant inffammation proteins between MD-D and HC; c & d. KEGG enrichment analysis of the significant inflammation proteins between MD-D and HC Figure 4a showed our nomogram prediction model. Figure 4b & c revealed the ROC curve of the nomogram with strong discrimination capability and an AUC of 1.000 and 95% CI (0.9404,1.000). The calibration curve showed effective calibration (p > 0.05) and decision curve analysis (Fig. 4d) demonstrated certain clinical benefits. Fig. 4A nomogram prediction model identifies major depressive disorder (MD-D) from health control (HC). a. The nomogram prediction model for MD-D; b. The receiver operating characteristic curves of the nomogram; c. The calibration curve of the nomogram; d. The decision curve analysis of the nomogram A nomogram prediction model identifies major depressive disorder (MD-D) from health control (HC). a. The nomogram prediction model for MD-D; b. The receiver operating characteristic curves of the nomogram; c. The calibration curve of the nomogram; d. The decision curve analysis of the nomogram MD-D vs. BD-D Ultimately, we identified two inflammation-related proteins with differential expression in the MD-D group compared to the BD-D group, both of which were up-regulated DEPs (Fig. 1a). Figure 5a displays a volcano plot of DEPs between the MD-D and BD-D groups, highlighting the top two increased DEPs. Two inflammation-associated proteins were found to be differentially expressed in the MD-D group compared to the BD-D group, and both were up-regulated (Fig. 1a). The volcano plot in Fig. 5a illustrated the DEPs between the MD-D and BD-D groups, with Fig. 5b showing the top two increased DEPs. ELISA confirmed that the expression levels of IL17-A and IFN-γ were significantly elevated (all p < 0.05, Fig. 5c). Fig. 5Comparison of the inflammation-related proteins between major depressive disorder (MD-D) and bipolar disorder depressive episodes (BD-D). a. Volcano plots showed significant differences of cytokines between MD-D and BD-D. Proteins highly expressed in MD-D were labeled in red. Proteins highly expressed in BD-D were labeled in blue; b. Box plot representing differentially expressed proteins between MD-D and BD-D; c. The expression levels of IL17-A and IFN-γ between MD-D and BD-D were shown in box plot by ELISA; d. Receiver operating characteristic curve analysis of IL17-A and IFN-γ between MD-D and BD-D. (* <0.05; ****<0.0001) Comparison of the inflammation-related proteins between major depressive disorder (MD-D) and bipolar disorder depressive episodes (BD-D). a. Volcano plots showed significant differences of cytokines between MD-D and BD-D. Proteins highly expressed in MD-D were labeled in red. Proteins highly expressed in BD-D were labeled in blue; b. Box plot representing differentially expressed proteins between MD-D and BD-D; c. The expression levels of IL17-A and IFN-γ between MD-D and BD-D were shown in box plot by ELISA; d. Receiver operating characteristic curve analysis of IL17-A and IFN-γ between MD-D and BD-D. (* <0.05; ****<0.0001) Moreover, ROC curve analysis indicated the AUC of IL17-A [p < 0.0001, AUC area = 0.846, 95%CI (0.747, 0.944)], and IFN-γ [[p < 0.0001, AUC area = 0.897, 95%CI (0.819, 0.974)] (Fig. 5d & Table 4). Based on the Youden index, their cut-off values were 3.645 pg/ml and 40.4 pg/ml, respectively (Table 4). The GO enrichment analysis also highlighted enrichment in the positive regulation of apoptotic processes and neutrophil apoptotic processes (Fig. 6a-b). Moreover, KEGG enrichment analysis indicated the IL-17 signaling pathways, TGF-β signaling pathways, HIF-α signaling pathways (Fig. 6c-d). Fig. 6Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) of the significant inflammation proteins between major depressive disorder (MD-D) and bipolar disorder depressive episodes (BD-D). a & b. GO terms of the signiffcant inffammation proteins between MD-D and BD-D; c & d. KEGG enrichment analysis of the signiffcant inffammation proteins between MD-D and BD-D Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) of the significant inflammation proteins between major depressive disorder (MD-D) and bipolar disorder depressive episodes (BD-D). a & b. GO terms of the signiffcant inffammation proteins between MD-D and BD-D; c & d. KEGG enrichment analysis of the signiffcant inffammation proteins between MD-D and BD-D Figure 7a presents the nomogram prediction model and Fig. 7b & c shows the ROC curve with high discrimination capability [AUC = 0.8833, 95%CI(0.7743, 0.9518)]. The nomogram’s calibration curve was effectively calibrated (p > 0.05), and decision curve analysis showed our model provided significant clinical benefits (Fig. 7d). Fig. 7A nomogram prediction model identifies major depressive disorder (MD-D) from bipolar disorder depressive episodes (BD-D). a. The nomogram prediction model for MD-D; b. The receiver operating characteristic curves of the nomogram; c. The calibration curve of the nomogram; d. The decision curve analysis of the nomogram A nomogram prediction model identifies major depressive disorder (MD-D) from bipolar disorder depressive episodes (BD-D). a. The nomogram prediction model for MD-D; b. The receiver operating characteristic curves of the nomogram; c. The calibration curve of the nomogram; d. The decision curve analysis of the nomogram Discussion The misdiagnosis of BD patients might lead to an elevated risk of mania and a postponement in receiving the correct diagnosis [7, 8]. Consequently, it is vital to distinguish between MD-D and BD-D for suitable intervention and the initial phases of treatment. Increasing studies have emphasized the role of the inflammatory response system as a crucial pathological factor [9–13]. Proteomics research has recently pinpointed serum and plasma proteins linked to mood disorders [19–21]. We are aware of only a limited number of studies that have used Olink Proteomics to conduct both quantitative and qualitative comparisons between MD-D participants and BD-D patients. So, we established a predictive model to pinpoint biomarkers that differentiate MD-D patients by Olink Proteomics. The research showed that IL17-A and IFN-γ may be the key biomarkers to distinguish MD-D from BD-D and we also identified their cut-off values. Our nomogram prediction model showed great alignment in the calibration curve and was demonstrated to be certain clinical value by decision curve analysis. More specifically, the AUC values of some biomarkers are very high, which may be related to the relatively small sample size. Thus, we need larger samples in future study for validation to avoid the potential risk of overfitting. IL-6 is a versatile cytokine that plays crucial roles in immune regulation and inflammatory processes. Previous findings revealed BD patients with a higher level of IL-6, whereas they did not increase notably in BD-D subjects compared with health individuals [22]. The CCL7 is primarily found in monocytes, fibroblasts, and T cells [23], and it draws monocytes and eosinophils, but not neutrophils [24]. Research conducted previously has revealed that MCP-3 is highly effective in determining the severity of diseases and predicting poor prognosis in both infectious and non-infectious diseases. Sun effectively distinguished between patients with mild and severe asthma through the use of MCP-3 [25]. Yang’s research showed a strong link between elevated serum MCP-3 levels and the severity of COVID-19 cases [26]. Transforming growth factor alpha (TGF-α) is a key cytokine involved in cell growth, differentiation, tissue healing, and the progression of diseases [27, 28]. TNFRSF9, also referred to as CD137, belongs to the TNF receptor superfamily and is primarily located on activated T cells, aiding in their growth and longevity [29]. It is also present in smaller amounts on other cell types, such as endothelial cells, macrophages, and monocytes [30]. Furthermore, several meta-analyses found that depressed patients have higher levels of C-reactive protein, IL-6, IL-10, IL-12, IL-13, IL-18, and TNF-α compared to healthy individuals [31, 32]. We used Olink Proteomics to compare inflammatory proteins between the MD-D and HC groups, finding that IL6, MCP-3, TGF-α, IL17-A, and TNFRSF9 were significantly higher in the MD-D group, as confirmed by ELISA. Moreover, Youden index revealed their cut-off values and our predictive model had certain clinical value. To ensure validation, future studies should include larger samples due to the relatively small sample size. IL-17, a pro-inflammatory cytokine from T helper 17 cells (Th-17), is vital for numerous immune functions, such as starting and managing immune responses by promoting the production of additional cytokines and chemokines [33]. Regarding psychiatric disorders, IL-17 has been a topic of considerable interest lately, although results frequently appear contradictory. According to a recent meta-analysis, this cytokine may not have a significant involvement in the pathophysiology of schizophrenia [34]. Elevated IL-17, among various cytokines, has been connected to post-traumatic stress disorder in those with rheumatoid arthritis [35]. Our research revealed that MD-D patients had elevated IL-17 A levels compared to HC or BD-D individuals. We also determined the IL-17 A threshold values to distinguish MD-D from HC or BD-D subjects, which were confirmed using a predictive model with clinical practicality. Moreover, KEGG enrichment analysis indicated the IL-17 signaling pathways in MD-D patients in comparison with HC or BD-D individuals. After reviewing related research, NFκB and p38MAPK signaling pathways may be the potential mechanisms in how IL-17 affect the mood disorder [36]. However, the specific mechanisms linking IL-17 A to mood disorders will be further studied in the future. Interferons (IFNs) are a group of cytokines produced by host cells when they encounter viral or bacterial infections. This family includes type I IFNs (α, β, ε, and ω), type II IFN (γ), and type III IFN (λ). Leukocytes, including innate immune cells like natural killer cells and macrophages, as well as adaptive immune cells, can release Type II IFN after being stimulated by cytokines such as IL-12, IL-15, and IL-18 [37]. The activities of IFN-γ include improving antigen display on MHC molecules, stimulating immune cells, and causing either cell death or proliferation [38]. Our study found that patients with MD-D exhibited higher IFN-γ levels than those with BD-D. We also established IFN-γ threshold values to differentiate MD-D from BD-D subjects, validated by a predictive model in clinical practice. We also need larger samples in future research for validation. Despite being the initial study to reveal the differential proteins and determine their cut-off values in MD-D patients versus HC and BD-D subjects, certain limitations should be acknowledged. First, we studied MD-D patients who did not exhibit manic symptoms for two years, and the shorter follow-up period may have influenced the results. Second, the accuracy of the results, such as the potential risk of overfitting, might be affected by the limited number of samples, so a larger sample size is necessary to confirm these findings. Third, we only chose 92 proteins related to inflammation and did not gather other possible inflammation biomarkers in this study because of technical limitations. Fourth, we did not perform external validation because of the small sample size. In summary, our results indicated that plasma inflammatory proteins can differentiate MD-D patients from BD-D, potentially serving as biomarkers for MD-D. In addition, we identified that their cut-off values. We developed a nomogram prediction model to effectively differentiate MD-D from BD-D patients, providing clinicians with a new method to swiftly identify MD-D and initiate early targeted therapies. We require larger samples to further validate our results.